What's new in Spitfire
The features of the recent releases; each card opens its guide or page. Your installation tells you about a new release itself: admins see the update command for that installation at the top of the UI. The command backs up the database first and keeps the settings (Updating).
One Spitfire Collector: server metrics and APM togetherServer metrics and APM traces now come with one install and one service; server metrics need no Tower. In the install dialog you choose the service names, the processes to leave out and the modules to turn off. Services found go to the Review list and the ones you approve are monitored. When the old Server Collector is installed, the new install takes it over: its name, interval and process count carry over and the old service is removed; each server shows as one record linked to its Collector. The Windows setup.exe/.msi and macOS dmg installers bring the new install too. Servers with metrics went up as well: Starter 5, Growth 20, Scale 40.All integrations on one pageNotification channels, data sources, connections and observability are now on the All integrations page. The Add button offers every kind in the same dialog and you switch between card and table views; a connection card shows its description, the production mark, the SSH tunnel and the names of its stored secrets, and the Test button shows the result right under the card. The old addresses redirect to this page, and Integrations in the admin menu opens it directly.Docs rewrittenThe docs pages are shorter and plainer. Every install ends in numbered steps, copy-and-paste commands and a success check that shows it works; the commands give the same result whichever folder you run them from. Prices are now only on the buy page.Follow your requests from inside SpitfireRequests you open from Spitfire no longer disappear. The My requests box next to the Request / idea button in the top bar shows a red count when the support team replies or a request's status changes. Open a request to read the replies and status changes in order and answer from the same place; close it when the problem is solved and reopen it within 30 days. Each user sees their own requests, installation admins see every request on the license. The same requests are on the Support page of your account on the site too.Install the APM Collector from a setup fileThe APM Collector now installs with a setup.exe or .msi on Windows and with an Apple-signed and notarized dmg on macOS. In the install dialog pick the Installer tab, download the file and copy the connection code; the installer reads the code from the clipboard and finishes with administrator approval. The msi and exe also install silently for software distribution. Removing it is just as easy: Installed apps on Windows, the Uninstall button in the dmg on macOS; every setting added to applications is taken back and no files are left behind.JavaScript in your tests: the script stepWhen the built-in extracts and checks are not enough, you now write JavaScript in the test editor. The script step reads and computes variables without sending a request; each step's Before script can change the request, for example sign it, and its After script checks the response with your own rule. vars, response, check, skip, log, hash and HMAC helpers and part of Postman's pm.* commands are ready to use. The Try button runs the code with sample values and shows the changed variables, checks and logs; code runs on the runner in a time-limited sandbox with no network or file access. On every edition, Free included.Ready for the campaign? A ready-made templateThe new E-commerce / Campaign readiness template runs a spike from 2 to 100 VUs, then 10 minutes of a realistic traffic mix: home, product, search, cart, login, order. It catches duplicate orders with two adds to the same cart at once and an order sent twice with the same Idempotency-Key. When the payment address does not look like a sandbox it asks once more before creating the test. A synthetic monitoring flow without payment is in the templates too.See at a glance which events notify youThe event list in the Add channel dialog is now called Events to notify on. Each row says plainly what the event means, the number selected shows on top, and Select all and None change it in one click. A new Slack, Teams, e-mail or webhook channel starts with every event ticked and you remove the ones you don't want; SMS, voice call, on-call and issue tracking channels start with the critical events only. Every name in the Systems you can connect list on the Integrations page is now clickable: a notification channel opens its add dialog, a monitoring system opens the add connection screen.Kubernetes tiles open the list right awayOn APM's Kubernetes screen, clicking a tile such as workloads not ready or pods not running now opens the list right under the tiles and scrolls to it; you no longer look for it below the filters and charts. In the table of hosts a service runs on, the columns are now CPU and Memory, so the host names get the room and long names read in full.Remove anything you installed, right awayThe removal commands now sit right beside the install: in the Add Tower dialog, the Install the server agent dialog, on each Tower's and Collector's page and on the Runners page there are copy-and-paste commands for Linux, Windows, Docker, Helm and Kubernetes, each ending with OK. The new Removing what you installed page puts everything in one table, from Spitfire itself to runners, Towers, Collectors, OTel agents and the Recorder extension; the database, queue and cloud pages have steps that undo the monitoring users and grants. On Windows, Defender no longer blocks the Tower and the agent, and on a Turkish Windows the scripts no longer ask again for the saved settings.APM on Kubernetes without an image changeThe new Kubernetes (operator, no image change) environment in the Add a service wizard connects Java, .NET 8–10, Python and Node.js services through the OpenTelemetry Operator: the agent comes with an init container when the pod starts, so there is no image to rebuild. The wizard gives five copy-and-paste steps, each ending with OK: cert-manager and the Operator are installed at pinned versions when missing, the token goes into a Secret, the address, sampling and service name go into an Instrumentation resource, the Deployment gets an annotation and its pods restart, and a check confirms the agent was added. Spitfire does not connect to the cluster; you run the commands.APM data now kept for monthsThe longest retention of traces, logs and APM metrics went from 3 to 30 days on Starter, 7 to 60 on Growth, 15 to 180 on Scale and 30 to 365 on Enterprise. No new key is needed; the new days apply with the update. So that no disk fills up on its own, an installation that never chose keeps at most 30 days; you choose longer in the APM settings. The APM time picker now has Last 30 days and Custom range: pick a start and an end to open any period of up to 31 days; the range stays in the address, so you can share it as a link.Service flow map and infrastructure mapClick a service in the APM topology and it moves to the middle: its callers on the left, what it calls on the right, step by step. +N opens a node's hidden neighbours in place and − closes them; the map lives in the address, so a link shares it. The new Infrastructure map shows hosts, the load balancers in front of them and the data tier on one screen: a host's colour comes from the highest of CPU, memory and disk, load balancer → member lines from the member's state, host → host lines from APM calls (Starter and up).Load balancer members and requests held at the LBHAProxy, Nginx, NetScaler, Kubernetes Service/Ingress and AWS ALB/NLB members are read and each member is matched to a host; pick the host yourself for one that does not match. The server page shows which pools the host is a member of. On NetScaler, the requests the load balancer counted are compared with what APM saw on the hosts behind it; when they differ, a Held at load balancer (estimate) badge shows with its evidence.Connected log stores in one searchAPM → Logs searches Loki, Elasticsearch, OpenSearch, Graylog, Splunk, Datadog and New Relic next to Spitfire's own logs; records merge into one list by time and each row carries its source's badge. Filters are translated into each store's own query language; a source that does not answer is named and the others are still listed. Stores are read read-only and records are not copied into Spitfire.Business APMDefine business metrics from the span and log attributes your services already send: order count, amount, split by payment method. The dashboard shows the value, the change against the previous period, failed records and the estimated lost amount. Funnels show conversion step by step and the most common error; business alerts catch drops and anomalies. Attributes that look personal cannot be chosen; nothing changes in the application.End-user depth: pages, crash groups, symbol filesLoad times, Web Vitals, LCP elements and INP targets for every page and screen; every session is marked phone, tablet or desktop. Error and crash groups split by version, device and country. Upload source maps and Android mapping files and the stack turns back into source file, line and function (Starter and up). A world map colours countries by sessions.Every plan's APM limits on one cardThe new APM limits card on the License page shows your plan's APM data per day (free 1 GB, Starter 5, Growth 25, Scale 100 GB, Enterprise unlimited), the traced services and web/mobile apps, request/response capture, connected log-store search, alert rules, SLOs and retention days. At a limit your applications are not affected; an upgrade applies the new limits at once.Update all outdated in one clickThe Update all outdated button on the Collectors page updates the fleet's outdated Collectors, Towers and APM Collectors together; Collectors first, then the Tower behind them. Towers installed with Docker or Helm are listed with a copyable update command. The Windows server monitoring setup was also tested end to end on a real Windows 11 in this release, and the bugs found were fixed.Noise reduction and probable root causeThe new Noise reduction page shows the funnel from incoming records to notices and the reduction rate; suggestions open a policy draft in one click and no policy changes by itself. A notice delay sends nothing for short-lived problems, maintenance windows hold notices until the window ends, and storm protection gathers a flood of events from one source into one event. The Probable root cause (estimate) card on the event page ranks the group's events by first opened, APM topology, layer, recent deploys and host saturation, with the reason for each; when the cause is clear, notices carry that line. The ranking is rule-based, no AI.Request and response in the trace, maskedThe span panel shows the method, URL, headers and body. Critical data such as national ID numbers, card numbers, CVV, tokens and passwords are masked before they are stored; add your own rules on the Masking page. Trace search finds free text in the URL, headers and body; a search by a raw value finds nothing.APM top lists, filters and sortingTop lists show the top web requests, SQL statements, errors and method hot spots, each with who calls it. One filter bar narrows by service, host, version, endpoint, k8s namespace / pod / node, container, status and exception with = or ≠ and suggests values with their counts; the filter lives in the address, so a link saves it. Every column sorts both ways.Host detail, service placement and remote updateThe server page shows IP addresses, the OS, CPU model, cores, memory and uptime; its tables sort and it refreshes live. The service page's Where it runs tab shows each host, pod and container the service runs on and with which version. Towers and APM Collectors update from Spitfire: the signed release is downloaded, and if it does not start healthy it rolls back. Each Tower's page shows CPU, memory, traffic, update history and a live log.NATS monitoring; Tomcat, JMX, Kafka Connect and more out of BetaNATS is now a data source: connections, slow consumers, message rates, JetStream and consumer pending. InfluxDB, ClickHouse, Cassandra, Tomcat, Java / JMX, Kafka Connect and NATS were tested end to end on several real versions and are out of Beta. If you don't have a receiver key, the I don't have the key link asks for your password and e-mails the key to you.Integrations tested one by one on real serversDatabases, queues, web and Java servers, gateways, search, log shipping, Kubernetes, Docker, SNMP and the Collector languages were set up on real instances, the guides followed line by line, and every bug found was fixed with a test. Setup scripts now also work with servers in a Docker container; Redis, Kafka and Couchbase take a private CA; errors say the cause and what to do instead of raw curl output; the Kafka, Redis, ActiveMQ, Couchbase and Memcached findings fire right in the real cases.The mobile app on Spitfire CloudOn cloud spaces mobile notifications are on from the start and go straight to Spitfire's relay, encrypted so only the phone can read them. No gate or relay URL to choose; the admin only turns them on or off. The new Connect your phone card on the Mobile & Gate page gives the server address to type into the app, with a copy button, and on the cloud the space name: in the app choose Add server → Spitfire Cloud and type it. On your own server the new SPITFIRE_PUBLIC_URL variable builds notification, e-mail and webhook links with the right address.Bank pack: traditional WebSphere, App Connect Enterprise, DataPower (Beta)Three new IBM sources under Infrastructure sources → Add, all read by the Tower itself: no Collector, and no setting changes on the IBM product. On WAS 8.5.5 / 9.0 the PerfServlet on one server gives the whole cell's PMI data: hung threads, threads waiting for a JDBC connection, full pools. On ACE 12 / 13, stopped integration servers and flows, error rate and slow flows; on DataPower, services and objects down and expiring certificates. was.sh, ace.sh and datapower.sh try the connection and add the source in one line.IBM MQ (Beta)One source is one queue manager: the Tower reads the mqweb REST API with a user in the MQWebAdminRO role, which may only run DISPLAY; no message is read. Queue manager state (active/standby), queue depths (dead letter and transmission queues first), readers/writers and channel status on the boards; no active queue manager, a retrying channel, a full queue or one with no reader and a growing DLQ become findings. MQ 9.2 and later.IBM Db2 (Beta)A new data source for Db2 LUW 11.5 and 12: the Tower signs in over the Db2 port and reads the MON_GET_* table functions once a minute; no Db2 client or driver to install. Panels for lock waits, deadlocks, buffer pool hit ratio, log and tablespace usage, backup age and HADR; the live view shows the top SQL by elapsed time and by CPU (literal values masked). The user gets only CONNECT and EXECUTE on nine table functions.OpenShift view (Beta)When the Kubernetes source sees the cluster is OpenShift, the cluster view gains an OpenShift part: ClusterVersion and updates, ClusterOperators, Routes and the days left on their certificates, Projects, DeploymentConfigs, builds, MachineConfigPools and, when allowed, the firing alerts. The Tower reads it with a read-only service account token; kubernetes.sh sees OpenShift and grants the access itself, and nothing on the cluster changes.Oracle RAC and ASM in the live viewWhen an Oracle source is a RAC, the live view reads the cluster through the GV$ views: each instance's state and sessions, gc blocks lost on the interconnect and the instances each service runs on. An instance down is a critical finding; sessions skewed to one instance and a service off its preferred instances are findings too. ASM disk groups show on a single instance as well. On existing installs, run the grant's RAC lines once.Built-in alerts: every service, no rule to writeThe new Built-in alerts section on APM → Alerts and anomalies → Alerts watches every service, new ones included: error rate (warning 5%, critical 10%), p95 latency (warning 1000 ms, critical 3000 ms), traffic stopped, traffic spike, new and reopened error groups. A minimum requests guard keeps quiet services from alerting at night. They are on in every edition, Free included, and do not count against the alert rule limit; switch any off, change its thresholds or leave services out. A rule you wrote on the same metric wins, so no double alerts.Collector naming rulesWhen only some of a server's apps belong together (the IIS sites abc, abc-payment and abc-order as abc, the other sites under their own names), open the server on App Collectors and press Naming rules. Rules are tried from the top and the first match gives the service.name: a glob (abc*) or, when ticked, a regex (^shop-(\w+)-v\d+$ → shop-$1). A preview shows the name each app would get and the rule that gives it; the Collector needs no update.CloudWatch metric pickerPick from CloudWatch, under an AWS source's Custom CloudWatch metrics field, lists the namespaces in a region, then the chosen namespace's metrics with their dimension names and series counts; pick a metric and a statistic and press Add. A saved AWS source's metrics can now be edited and saved from the card at the top of its Dashboard. The list only reads and the policy does not change; the picker needs Tower 1.4.14.License key from a file and from the command lineLoad from file on the License page reads the .lic file from the e-mail and installs the key at once. Invisible characters added by mail clients, line breaks, quote markers and words before the key no longer break it. When the browser can't install it, spitfire-controller license install on the server does it with the same checks. Instead of a raw "Failed to fetch", the page tells an unreachable server, a timeout and a proxy answer apart, each with what to do.Copy buttons work on a plain-http install tooOn Spitfire opened at an address without https the browser gives no clipboard access, so Copy buttons (the two-step verification recovery codes among them) did nothing. Every Copy button now says Copied when it worked; when the browser still refuses, the text shows selected in a read-only box and Ctrl+C (Cmd+C on a Mac) does it. The recovery codes can also be saved as a .txt with Download.Requests and ideas without leaving SpitfireRequest / idea in the top bar opens a form inside Spitfire: pick the type (Request, Suggestion, Bug), write it and attach up to 3 files if you like. The Report button on error boxes fills the same form in with the error; Preview shows the whole text that will go. Send opens the ticket on Desktik, and the form shows the ticket number and a Follow on Desktik link. When the server cannot reach the internet the text is not lost: copy it, e-mail it or press Send from the site. Nothing is sent by itself.Sweeper: page cap by planWithout the Sweeper module a sweep now opens at most 50 pages on the free edition, 100 on Starter, 500 on Growth and 1,000 on Scale (it was 50 on every edition). The cap comes from the plan, so existing license keys get it without being reissued. The Sweeper module (an add-on, its 7-day trial or Enterprise) allows sweeps of up to 5000 pages and scheduled sweeps.On-call channels: PagerDuty, Opsgenie, Alertmanager; Google Chat and WebexAdmin → Integrations → Add channel has a new On-call group. PagerDuty opens an incident when a problem starts and closes the same incident when it clears; Opsgenie does the same with an alert, and Alertmanager sends a firing alert and resolves it. Google Chat posts a card to a space, Cisco Webex a message to a room. All are free on every edition.ArgoCD and Jenkins deploy marksWebhook inputs take ArgoCD and Jenkins; the dialog gives the ConfigMap and the Jenkinsfile with your token filled in. Each deploy becomes a vertical mark on the APM charts (red when it failed), listed below with its revision and who ran it. Why did it break? ranks the deploy first when errors rose right after it.Google Cloud Monitoring, in depthA metric explorer on the Google Cloud source's board: chart any metric type (custom, workload, prometheus…) with filters and grouping, or with PromQL, then save the query and pin its chart. Uptime checks show with their result and latency per region; open alert incidents land in Events and go through the event policies.New services on AWS and AzureThe AWS source also reads ElastiCache, SNS, API Gateway, MSK, S3, CloudFront, Kinesis and Fargate (ECS Container Insights); the Azure source Service Bus, Event Hubs, Application Gateway, Redis, Cosmos DB, Key Vault and Functions. On AWS you can add your own CloudWatch metrics too (up to 20). On existing sources, add the new services to the Services field.IBM Cloud (Beta)A new cloud account type: the Tower reads IBM Cloud Monitoring with a read-only service ID. VPC servers and load balancers, IKS / OpenShift, Databases, Event Streams and Cloud Object Storage come with a ready dashboard and findings. The quick setup command creates the service ID and its access itself.Kong, Envoy, Traefik monitoring (Beta)A new API gateway / proxy group under Infrastructure: boards for requests, 5xx, p95 latency, connections and upstream health; findings for unhealthy targets, open circuit breakers, rejected config and TLS certificates about to expire. Spitfire only reads the metrics endpoint and never touches your config.Hazelcast, Kafka Connect, Keycloak, Vault (Beta)Four new sources the Tower reads itself, with no Collector config. Cluster and member loss on Hazelcast, failed tasks on Kafka Connect, failed logins on Keycloak, seal status and audit errors on Vault. Keycloak and Vault sit in the new Identity and security group.Apache HBase (Beta)The Tower reads the Master's and RegionServers' JMX JSON itself; no Collector, nothing changes in HBase. Boards for RegionServers, RIT, read and write rates, block cache, the compaction queue and heap; a dead RegionServer is a critical finding.APM guide for C++ servicesSend traces with opentelemetry-cpp and the OTLP HTTP exporter: one header, two lines in main, a server span that continues the incoming traceparent and three lines in CMakeLists.txt. C++ can be picked as a language in the Add service wizard too.Every app on a server under one service nameTo see apps such as the IIS sites abc, abc-payment and abc-order as one service in APM, open the server on App Collectors, type the name in the Send every app under one service name card and press Apply. All of them, the ones auto-trace switched on too, send under that name; each app's own name stays as spitfire.app, so you filter by it on the service page, on Top requests and in the trace search. A name typed in an app's own dialog wins. The Collector restarts apps only if it was installed with restart consent.Sweeper: changed only or full scanChanged only opens again the pages whose content, ETag or linked files (scripts, stylesheets, images) changed since the last full scan; the others keep that scan's result, and the run page says which is which and from when. Every 7th scan, or once the last full scan is older than 7 days, runs as a full scan. Pick the type in the sweep's window or from the arrow next to Start; in CI, --scan changed.Menu on the leftA module's pages now sit in a column on the left, next to the module rail; Collapse the menu shrinks it to icons. If you prefer the old layout, Move the menu to the top at the foot of the column brings back the tab row, and Move the menu to the left returns. The choice is kept in this browser, for your user.Spitfire Cloud: nothing to install, a free spaceSpitfire now also runs on our servers, as a separate workspace for each customer, with its own database, keys and resource limits. Sign up and press Open my free space; your company.app.usespitfire.com address opens and the first admin opens their account from the invite link. The free space stays and keeps the last 7 days of data; buy a plan and the same space goes on. Nothing changes in a Spitfire you install on your own server.Target domain verificationIn cloud workspaces full load goes only to domains you have proven you own: on Tests → Target domains add the domain, put the token in a DNS TXT record or a /.well-known/spitfire-verify file and press Verify; subdomains are covered too. A run against an unverified target is capped at 50 VUs and 2 minutes. In your own installation this check is off.Log collectors under Add connectionAPM → Add connection lists Log collectors right after the log stores: Syslog, NetScaler, log files, journald, Windows event logs, Kubernetes, Fluentd/Fluent Bit, Splunk HEC, Logstash and Vector. A card opens that source's Collector setup: options, configuration, one-line install, test record and first records.The installer offers Docker and k3sOn Linux without a Kubernetes cluster, install.sh offers to install a single-node k3s (it asks first, default no); without Docker it installs or starts it with your consent. The commands are printed before each question. For CI and automation there are options that do it without asking; --yes alone installs nothing.Intro tour and Request / ideaOn the first visit a short tour shows the menu and the main pages; Show the intro again in the user menu opens it again. The Request / idea link in the top bar writes your request straight into our support system.End-user monitoring: from the page and the tap down to the databaseData from your websites and mobile apps on APM → End users: first paint and Web Vitals (with good/needs-work/poor bands), page and resource timings, slow pages, JS errors and mobile crash/ANR groups, cold start; device brand/model, OS, browser, country/region/city. Each session's timeline shows the steps before a crash. A request that starts from a page or a tap is one trace down to the backend and the DB query; the trace page says where it started. Copy-paste guides for HTML, React, React Native, Flutter, Android and iOS. IP addresses are not kept; query values in addresses are masked.APM Collector: one command, and it finds your apps by itselfAPM → App Collectors → Install Collector: the wizard gives one command, or on Windows and macOS an installer and a connection code if you prefer. The Collector finds the Java, .NET, .NET Framework, Node.js, Python and PHP apps on Linux, Windows and macOS by itself and switches auto-trace on in the apps it finds; it restarts nothing without your consent. Switch tracing on and off per app on the same page and the traces flow into APM. It uses the official OpenTelemetry kits, opens no port and only talks to Spitfire. For apps in Docker and Kubernetes it shows a settings block to copy.APM overview and the TV screenAPM's first page is now the overview: every service and every host is a card. A service card shows requests/s, p95, the error rate and a line of requests; a host card CPU, memory and the fullest disk. Red cards come first, then orange and green; under an orange or red card the reason is one line. Click a card and the service or host page opens. The Wall screen button opens the cards full screen without the menu; add ?tv=1 to the address to open it straight on a TV.Anomaly detection: no rules to writeSpitfire learns each service's requests/min, average latency, p95 and error rate from the same times of day over the last 7 days. When a value rises or drops more than 15% and stays outside the learned band for 5 minutes, it opens an event, fires an alert and notifies the owner and channels; traffic stopping altogether is critical. It closes by itself when things are back to normal. It starts on its own after at least 4 learned days; the settings are on the service page's Anomaly tab. The new apm.anomaly event comes ticked on notification channels (except SMS/IVR and trackers).Manage the Collector from the panelThe App Collectors page shows each Collector's state, CPU/RAM and live log. Change its log level (optionally back to info after a while), stop it and start it again. No need to log on to the server: the Collector opens no port, and a change reaches it with its next report.DB queriesAPM → DB queries lists your services' database queries normalized (values become "?", nothing is kept) with total time, calls, latency, errors and rows. The slowest calls open their trace. Also a tab on the service page.Top requests and errorsTop requests puts the endpoints of all services in one list: requests, total time, average, p95 and error rate; total time tells where the time goes. Top errors ranks the most frequent error groups with their services and endpoints.Trace filters and groupingTrace search gets filters on the left by service, status, status code, host, environment and duration bucket, with a count beside every value. Group the list by service or operation: requests, error %, average and p95 on each row. The link keeps the ticks and the grouping.Kubernetes: every selection opens its own listClick the pod count for the pods of every namespace, a node for the pods on it, a namespace for its workloads and pods, a Deployment, StatefulSet, DaemonSet or Job for its pods only. The lists filter by namespace, workload, node, phase, ready and restarts. APM → Kubernetes and the data source's Dashboard are one screen; the selection stays in the address.New menu: a module rail and tabsA slim module rail on the left (APM, Load Testing, Admin, Help) and the chosen module's tabs on top. No second side menu; the whole screen is left for content. Same colours.Integrations in one placeConnections, APM, notification channels and data sources under Admin → Integrations in one list, each with its status, as cards or a table. Test, edit and delete from the same place.Alerts against the usual hourTwo new APM alert rules: they warn when p95 latency or the error rate climbs unusually high against the same hour of the last 7, 14 or 28 days. A service that is calm at night and busy at noon gets its own usual instead of one threshold. The rule dialog shows a live preview.Installs check firstThe command, the Windows .exe/.msi, the macOS installer and Helm check Docker or Kubernetes before doing anything, and install or start Docker with your consent. Windows works without Docker Desktop too: Docker Engine is installed inside WSL2. When the install ends, the admin code and addresses stay in the window. If something goes wrong, and only with your consent, a redacted error report goes to support.Confirm dialogs no longer get stuckAnswering a confirm dialog such as Not tested when saving a connection or Trust the host key? on an SSH tunnel very quickly while it was opening (pressing Enter twice, say) left it invisible on the page, and the buttons of the form below stopped taking clicks. Confirm dialogs now open and close without the fade. In the audit log, the Show every workspace's entries box ticks at once.Kubernetes: from a count to the pod, "Why did it restart?"The Kubernetes data source page shows the cluster first: workloads not ready, pods not running and the pods with the most restarts. Every count in the cluster summary is clickable: a list first, then the detail. A deployment shows how many pods are up and why the others are not (CrashLoopBackOff, ImagePullBackOff, Pending, Unscheduled). A Why did it restart? panel for every pod and container: the reason, the restarts in the last hour with their times, the peak memory before the restart against the limit, Kubernetes events and log links. Chart legends are one row with short names.Preparation guides: Kubernetes, OTLP, PostgreSQL and SSH tunnelConnecting Kubernetes: Tower runs inside the cluster and opens a single outbound connection, no port is opened into the cluster; the pods meet Pod Security restricted, and Secrets are entered at a hidden prompt, so they stay out of the shell history and the manifest. The OTLP sender guide opens from the APM page: a verify command before you send, Test showing the first request's arrival (its time and source IP), and the receiver row with the last data per signal and the rejected requests. PostgreSQL and SSH tunnel: a monitoring role that cannot touch the data, generating the tunnel key in Spitfire, preparation commands filled from the form's server and port, and a layered connection test. Every step is copy and paste, with a success check at the end.429 is not an error: the rate limit findingThe target's 429 (Too Many Requests) responses no longer count in the error rate, the thresholds or the capacity test breaking point; they still show in the status code breakdown. The report has a This is not an error finding: the steps that got 429 and how many, the second of the first 429 with the VUs and requests/s at that moment, the limit headers the server sent (Retry-After, X-RateLimit-*, RateLimit) and the Spitfire outbound IPs and X-Spitfire-Run header to have allowed by the target's admin. The capacity card shows each step's 429 count in its own column.Test from an API doc: page address and variablesWhen importing, a docs page address works instead of a JSON file (e.g. https://…/api/docs): Spitfire finds the OpenAPI file on the page, or shows the addresses it tried. Path variables such as /orders/{orderId} are added to the test variables, with the doc's example value when there is one; the editor says the variables were added and you fill in the values, and Go to variables takes you to the empty field. In the test editor each step is its own white card: thicker border, number badge and space between steps.APM live refresh, failed spans in redA time range picker and a refresh button at the top right of every APM page: Off, 5s, 10s, 30s, 1m, 5m, 15m; 1m by default on relative ranges. The choice stays in the link, so a shared link opens with the same setting. Refresh waits while the tab is hidden, a dialog is open or you are typing. Failed spans are red in a trace and the Errors: N badge jumps to the first one. Topology boxes are wider, and a legend explains the colours and lines. APM Services shows recommended tests as one line above the table with a count badge on each service; Show all opens a searchable list grouped by service.Filtering tablesOne toolbar on the main tables (tests, runs, monitors, connections, data sources, agents, users, API tokens, audit log, event policies, SLOs): search that knows Turkish letters (istanbul also finds İSTANBUL), per-column value filters with row counts, date and number ranges, chips with Clear, and N / M shown. Filters stay in the address and / jumps to the search box. Events, Logs and APM Errors share one filter layout: a filter column you can hide, a full-width list and Clear all; logs are one line each and open in full on click.Event policy templatesSearchable, categorised Start from a template cards in the New policy dialog. Open policy counts on the tabs and in the menu, and a link to the built-in policies.Automatic updateOn Docker installations with one-click update on, when an installation admin turns on Update automatically when a new release is out under License → Version and updates, a new release installs without waiting for a click. The signature check, waiting for runs and the database backup still happen; every update is written to the audit log. Off by default.Copy beside the command, a separate source catalogCopy buttons sit beside the command instead of on top of it: Kubernetes, Docker, IIS, SNMP, update commands, CI snippets and the Prometheus config. In Infrastructure monitoring the source list and the Add source catalog are separate pages. NATS brand icon on connection type badges. The Turkish screens and docs now say Audit log.Incident management with DesktikPick Desktik in an event policy's Tracker action and Desktik opens one ticket per event; ack, resolve, reopen and severity changes go to the same ticket. The channel works on Desktik's events API: turn on Settings → Monitoring events in Desktik, and an events-only Desktik key (dtk_…) is enough, no Spitfire token needed. Acks and resolves done in Desktik are read every 30 seconds and applied to the Spitfire event. Sending to the desktik.com cloud is off by default; when on, only the ticked fields go (Title, Service and Severity by default). A Desktik on your own server gets every field. The Desktik channel is free in every edition, Free included.Pasted values are trimmedForms trim spaces, line breaks and invisible characters (zero-width space, BOM, no-break space) pasted around values: connections, SSH tunnel, APM, data sources, notification channels, SMTP, OTLP, SSO/LDAP/SAML and SIEM streams. Values saved earlier are trimmed when used too. Passwords are kept as typed.Scrolling dialogs, white badges, log collectors in the pickerOn short screens the connection, policy, alert, SLO and Sweeper dialogs scroll again, so Save and Cancel are in reach; the kind picker's category menu scrolls on its own when it is taller than the window. Type badges are white with brand-coloured icons: Connections, test steps, dry run, APM and notification channels. APM's log collector sources use the new kind picker too: category menu, search, icon cards.Roles: Admin, Editor, Viewer and your ownBuilt-in Admin, Editor and Viewer (read-only) roles in every workspace. The installation admin defines custom roles under Infrastructure → Roles: None, Read or Edit for each area, such as Tests, Runs and reports, APM, Events, Channels, Monitors, API tokens and the Audit log. An area set to None leaves the menu, Read turns its edit controls off; the API follows the same rule and a token never goes beyond its owner's role. Only the installation admin creates users and roles; workspace admins assign roles to members. With SSO, Role groups map directory groups to roles. Existing User accounts become Editors. In every edition, with no plan limit.Built-in policies and cross-source correlation31 built-in rules for Kubernetes, Docker, Linux, Windows and network (SNMP) come on with the installation: a server sending no data, CPU, memory, disk and inodes full, a stopped systemd unit or Windows service, a pod restarting, CrashLoop, OOMKilled, a node not ready, a full PVC, an interface down or saturated. Each can be switched off, its warning and critical thresholds and duration changed, and any number of channels picked; Back to defaults undoes it. Events gets a Source column and a Groups view: problems of the same host or service that come from different sources (built-in policies, monitoring connections, webhook inputs) within 10 minutes gather in one card and only the first is notified; the noise figure on top shows how many rows the list is down to. The installation admin cuts notices during maintenance with Stop all alarm notifications. Notices go out through a durable queue and a failed delivery is retried after 30 s, 2 min and 10 min. Built-in policies do not count towards the policy limit.A new kind picker, Test before saving, preparing ZabbixA new picker on APM, notification channels, data sources and connections: a category menu on the left, search and icon cards with a short description. Test in the connection dialogs before saving; every connection test shows the same result view. Query connections take extra HTTP headers and show the egress IPs Spitfire connects from. A new Zabbix preparation page, and "How to prepare <kind>" links in the dialogs.A simpler release banner and a Collectors pageThe new-release banner is simpler: what's new opens on the site and How to update shows only the commands; with one-click update on the banner has Update now, and on Docker with it off, a link to turn it on. Collectors in the APM menu for installation admins: status, machine details, live log and log path; the APM host page links to that host's Collector.Reports: a Brief, six formats, send to a channelEvery report opens with a Brief: the verdict (PASS, BORDERLINE, FAIL), what was measured, the change against the previous one, a likely cause only when there is evidence, labelled certain or likely, and what to do; it also shows as a card on ended run and comparison pages and in the SLA card. Run, comparison and monthly SLA reports download as HTML, PDF, Word (.docx), CSV, Markdown and JSON; ?format= works on the report addresses and share links too. The Report menu is the same everywhere: Download, Send to a channel…, Schedule…, Share link…. Send to a channel delivers the report to one or more channels: e-mail, Slack (with a bot token), the new Telegram channel and Jira get the files attached; Teams, other trackers, SMS and JSON webhooks get the brief and share links valid for 7 days that can be revoked. Each channel's outcome shows in one table.Scheduled reportsTest → Report schedules: a period report of a test, a group, a tag, the monitors or the monthly SLA goes to your channels on its own, daily, weekly or monthly, at the hour and time zone you pick, compared with the period before. Send now tries it without waiting; the list shows the next send and how the last one went. From the CLI: spitfire cloud reports schedules list, create, run and delete. Free 1, Starter 5, Growth and up unlimited.Tower never slows your servicesThe Tower writes nothing to disk any more: exports from Collectors are answered at once and wait only in a small in-memory queue, like the OpenTelemetry batch queue (64 MB by default). If the controller cannot be reached for 60 seconds the queue is cleared and new exports are dropped without retries; forwarding resumes by itself when the link is back. Settings --queue-mb (1–256 MB) and --drop-after; a new Queue column on the Towers page. Every Collector config Spitfire generates also comes with an in-memory, non-blocking sending queue and retries for at most 60 seconds.APM alerts and SLOsAlert rules with warning and critical thresholds on a service's (optionally an environment's and an endpoint's) p95, p99, p50, average duration, error rate, request rate drop/spike, or new or reopened error groups, evaluated every minute. Add alert on the service page fills the thresholds from the last 24 hours. The notice reaches your channels with the value, threshold, chart, slowest and failed traces and a Why did it break? link, and issue trackers get an issue. Silences and maintenance windows mute the notice, not the evaluation. SLOs: availability or latency, a 7, 28 or 30 day window, the error budget left and fast/slow burn rate alerts; Make an SLO from this p95 on a capacity run. In the CLI, --fail-on-firing exits 99 before a deploy. Free 3 rules + 1 SLO, Starter 20 + 5, Growth and up unlimited.APM › LogsEvery log that arrives over OTLP on one screen: search by text, trace id or span id, level, service, environment and host filters, key=value attribute filters. On the chart stacked by level, click a level to add it to the filter, drag to narrow the range. From a log row to the trace's waterfall with that span picked; a Logs tab on the service page and See logs on an error group. Patterns count logs of the same shape on one row, Live adds new logs every two seconds. Saved searches are shared in the workspace; CSV and JSON export up to 10,000 rows, each one in the audit log.Turn this search into an alertA search on the Logs screen becomes an alert rule: when more log lines than the threshold match the same filters within the minutes you set, or a new pattern shows up, a notice goes to the channels you pick, like the other APM alerts.EventsTest → Events gathers every event Spitfire raises (synthetic monitors, APM and SLO alerts, error groups, regressions, scheduled runs, connection health, Sweeper) and the alerts posted to webhook inputs in one list. While an event with the same key is open a repeat adds no row, it raises the count; a problem back within 10 minutes reopens the same event. Six severities from critical to ok; the sources' own levels are mapped to this scale. Status, severity, tag, source, service, host and environment filters; ack, resolve, notes and bulk actions, each one in the audit log. Your own events come in with POST /api/v1/events and an API token. Kept 7 days on Free, 30 on Starter, 90 on Growth, 365 on Scale and Enterprise.Event policiesOrdered rules run when an event opens: conditions on source, type, service, environment, host, severity, tags, text and a time window (outside business hours, night, weekend); actions to suppress, set the severity, enrich (owner, team, runbook), route, escalate without an ack, renotify, auto-resolve, open a tracker issue and call a webhook. Try runs the draft over the last 7 days of events without saving and shows each action's effect with samples; no notice goes out. Every save is a new version and History shows who changed what. Notices carry signed ack and resolve links that need no sign-in, last 7 days and work once. Free 3, Starter 20 policies, Growth and up unlimited.Events from monitoring connections, correlation and flapThe open problems of the Zabbix, Datadog, New Relic, Dynatrace and Instana connections are read once a minute and land in Events; a problem closed in the monitoring system resolves its event. It only reads, and notices go out through a policy's route. The correlate action groups events with the same values of the fields picked (by default service + environment, 10 minutes) under the first one; only the group's first is notified. The flap action marks an event that keeps opening and closing (by default 4 openings in 30 minutes) as flapping, sends no notice while it flaps and one notice once it stops. Try also shows how many events would join a group and how many would flap.Sweeper Interactive modeOn staging a sweep does more than read: it tries every page's buttons, menus, tabs, checkboxes, radios, selects and forms in the real browser, looks for a new console error, exception, 4xx/5xx, blank page or a click that missed its element after each try, and reports it with the element's label, selector and a screenshot. Requests to other hosts and sign-out are stopped in the browser; pay, send and log out buttons are never clicked. It cannot start on a production environment's host. 20 interactions a page without the module, 500 with the Sweeper module.Help when you get stuck on a formIn the test editor, the APM service wizard, schedules, connections, notification channels and when adding a monitored system, stop on a field for 20 seconds (5 after the same error twice) and a small bubble opens next to it: the field's tip, its section in the docs and an Ask box for the assistant. It never opens while you type and shows once per field per session; Don't show again turns it off. It runs in the browser and sends nothing.Write approval for schedules and comparisons tooThe approval for SQL and Mongo statements that change data now also opens when you save a schedule, monitor or comparison, and is asked per environment. In a bulk SQL script, transaction statements such as BEGIN, COMMIT and ROLLBACK no longer count as writes.Login & tokenDefine the login request once: every VU signs in, takes the token from JSON, XML, a header or a cookie and adds it to every request as Authorization: Bearer (or as a query parameter). It signs in again before expiry, on a timer you set or on a 401/403, and retries the request once. The login is measured in its own Login row. There is a Try the login button, a manual token mode stored encrypted, and public steps that get no token; OpenAPI and Postman imports move their auth into this block.The run's environment and topologyThe Environment tab on the run page shows the infrastructure under the test during the run: Kubernetes clusters (nodes, pods, pods not ready, restarts, CPU and memory average and peak), hosts, database, queue and cache sources, services and a map from the load generator to the services and on to DB/cache/MQ. It is stored after the run and goes into the HTML/PDF report as an Environment section. Spitfire installs nothing, it only reads; a source it cannot read shows why and never affects the run.E-mail the report, download WordReport → E-mail the report: your own address is filled in, up to 20 recipients, the PDF attached (Word or HTML on request) and an optional note; every send goes to the audit log. A test can e-mail its report when each run ends, or only when it fails. The report also downloads as Word (.docx). The Result card has a verdict badge and p95, p99, error rate and requests/s tiles coloured against the thresholds. Needs SMTP set up.The network side of the breaking pointThe breaking point card also shows the network side of the test's own requests: connect and TLS time, pool wait, the share of new connections and network errors by kind (refused, reset, timeout, DNS, TLS), against the stage before and each line with where to look. SNMP interface errors and NetScaler series are labelled network in the ranking.Findings written back to Datadog and New Relic tooAs with Zabbix, Dynatrace and Instana, test findings now go as events to Datadog (Events API v2 of the connection's site, warn/error, tagged source:spitfire) and to New Relic (the US or EU Event API, type SpitfireFinding by default; the License/Insert key is stored encrypted). The same categories and levels, and a Send test event button. Off by default.NetScaler and Citrix (Beta)Two new monitored systems under Infrastructure, read-only. NetScaler (Citrix ADC) is read over the NITRO API with a user allowed only stat/show: CPU, memory, throughput, HTTP, SSL, interfaces and every load balancing vserver with its members; a DOWN vserver, a DOWN member and response time rising with load become findings. Citrix Virtual Apps and Desktops (Delivery Controller) and Citrix DaaS (Citrix Cloud API) are read with the Read Only Administrator role: sessions, registered/unregistered machines, load index, logon duration by phase, connection failures. Both are Beta: not yet tried against a real NetScaler or Citrix site.APM: deeper traces and errorsThe span panel shows self time and its share of the trace, trace/span/parent IDs with copy, the parent and child spans, links to the host, service, error group and Kubernetes/Docker place, exceptions with their stacktrace and the span's logs. The waterfall folds and has Only errors, service colours, Critical path, Find a span, a Table view, keyboard navigation and a ?span= link. Errors get Regressed and Unhandled badges, Common to every occurrence, a stacktrace that highlights your own code and See traces; opt in to apm.error.new and apm.error.regressed to be notified. Every APM table sorts and searches.IVR voice calls; channels free on every planThe new IVR channel rings the on-call phone and reads the event out. You bring the provider: the Generic HTTP voice API (JSON) or Twilio Voice style preset fills the URL, the body and the header. By default it calls for critical events only (a failed run, a monitor going down), at most 5 calls an hour; Make a test call shows the provider's answer. E-mail, Slack, Teams, SMS, IVR and webhook channels are free and unlimited on every plan, Free included.APM: Kubernetes and Docker pagesAPM › Kubernetes goes from the cluster to nodes and namespaces, workloads, ReplicaSets, pods and containers: state, ready, restarts, age, CPU and memory against request and limit, network and charts. The APM services running in each pod and container are listed with requests/min, error % and p95. The tables sort and filter, the address keeps the selection, and Topology draws the calls or the placement on nodes. APM › Docker groups containers by Compose project; the service page has a pods card, and Services and Hosts link to the pod, node or container.Manage your agents in one placeInfrastructure › Agents lists every agent and Tower with its version, status, last report, points per second, failed and queued points; spitfire-agent supervises the collector. An agent's page has process CPU/memory charts, the log level (per agent or the fleet default; debug goes back to error by itself when its time is up) and a live log (filter, pause, copy, download the last N lines). Restart and Update now go to one agent or the selected ones after a confirm; if the new version does not start, the agent goes back to the old one. Auto-update is optional, with an update window. No remote shell: install admins only, every action in the audit log.Connection health and alertsAPM connections, event sources such as Zabbix, write-backs, a Tower's data sources and issue trackers such as Jira are recorded on every call as answering or not; the ones no call reached for a while are checked every 5 minutes. The lists show a Health badge, the last OK time and the last error (passwords and tokens hidden); a click on the badge opens the history. After 3 failed calls in a row (settable from 1 to 20) a connection is Down; it is notified once when it goes down and once when it is back, to the channels you pick or the workspace's channels. The APM, Infrastructure and Integrations pages open with a summary.Log collection and the Logs and load chartThe Log collectors card on APM installs an OpenTelemetry log collector with one line on Linux, Windows or Helm, for syslog, NetScaler, log files, Fluentd/Fluent Bit, Splunk HEC, journald, Windows event logs, Kubernetes, Logstash and Vector. The run window's logs are pulled read-only from Elasticsearch, OpenSearch, Graylog, Splunk, Datadog and New Relic. On the Backend tab the Logs and load chart draws error/warn/info records against requests per second with their correlation (r), next to the most frequent error and warn messages; when error logs rise with the load, the executive report gets a finding.Oracle and MongoDB: slow queries and locksOn Oracle, the top SQL from V$SQLSTATS by total time, mean time and calls, and lock waits from V$SESSION/V$LOCK; when a grant is missing, the grant script is shown. On MongoDB, the profiler's slow operations from the last 15 minutes (with a COLLSCAN badge), running operations from $currentOp and lock waits. Values are replaced by question marks and nothing is stored. Needs Tower 1.4.2.Schema suggestions: pick a count, reshuffle, MongoDBYou pick how many queries to suggest (10–100, default 20), from the tables you tick or from all of them; the suggestions are spread over tables and kinds, and Reshuffle makes a new pick. Besides PostgreSQL, MySQL, SQL Server and Oracle it now works on MongoDB connections, which are read with read-only commands only. Only inside your own Spitfire, for installation admins, with consent.Queries from a CSV and a strong consent for statements that change dataSQL and MongoDB steps take queries in bulk from a CSV: the query column is required, name, weight and params are optional. Every row is coloured by kind: inserts and updates amber, deletes, TRUNCATE and DDL red. A test with statements that change data opens a strong warning showing what changes on which connection; without the box and the confirm word no run, comparison, schedule or monitor starts. The consent goes to the database and the audit log.Logic comes over from JMeter tooA Loop Controller becomes a repeat block, If and While Controllers if and while blocks (a simple comparison, also inside __jexl3/__groovy), Switch, Random and percent Throughput Controllers a weighted choice block. A Module Controller copies in the controller it points at, an Include Controller the .jmx picked with the plan. The common JSR223/BeanShell lines (vars.put of a JSON value, setFailure, assert) become extractions and checks; the Boundary and CSS Selector Extractors convert too. Every line that cannot be converted is listed in the report.Loops, conditions and parallel requests from k6A for loop with constant bounds becomes a repeat block, if/else an if block (res.status, body, a JSON field, a header, a variable), Math.random() < p a choice block and http.batch a parallel block. A CSV read with SharedArray or papaparse becomes a data binding when you pick it with the script. Blocks show as badges on the step cards in the editor; bulk import matches the CSVs too.k6 script metricsThe script's own Trend, Counter, Rate and Gauge metrics (k6/metrics) are stored with the run. The Script metrics card on the run page shows the first 20 with type, count, value, p95 and max; a row opens the per-second chart. The HTML and PDF reports have the same table, and the REST API has them as custom:<type>:<name>.Postman pre-request scriptsOn import, the common lines of pre-request scripts become template values: fixed assignments, Date.now(), ISO time, uuid, random numbers, CryptoJS HMAC, MD5/SHA, btoa and adding a header. New template functions: {{$unix}}, {{$hmac sha256 key msg}}, {{$hash md5 x}}, {{$base64 x}}. Postman's {{$timestamp}} now becomes {{$unix}}. Lines not converted are listed.Extracting between boundaries and with a CSS selectorTwo new extraction sources: Between boundaries (the value between a left and a right text) and CSS selector (HTML). The JMeter import turns the Boundary Extractor and the CSS Selector Extractor into them.--fail-on-regressionWhen the thresholds pass but the run is worse than the baseline run (a step with ⚠️ in the PR summary, or requests/s down beyond the tolerance), spitfire cloud run exits 96 and the CI job fails. The GitHub Action and the GitLab component have a fail-on-regression input. With no baseline run it writes a note and exits 0.Add a runner in one clickAdd runner at the top right of the Runners page makes the token itself. The commands come ready with the token, the controller address and the location, each with a copy button; the dialog tells you when the runner has joined.Monitors and data bindings from an empty pageThe empty Synthetic monitors page has an Add monitor button and a link to pick a test, which opens with the monitoring dialog ready. The Data files page explains what a data file is for; Bind to a test opens the editor with the file bound and its {{name.column}} variables ready to copy.The integrations catalogThe Integrations page has four panels: notification channels, issue tracking, open standards and monitoring events. Each panel has one In your plan or Add-on badge, one upgrade button and the set-up link. Trial buttons name the module or add-on you are trying, and no trial banner shows when your license already includes it.Links everywhereWherever text sends you to another page, tab or card, the name is now a link, and it opens according to your role. Tables that show a test, run, runner, connection, data file, user, group, workspace or Tower link to it. Long explanations sit behind the orange i icon; buttons that read a document show a spinner.Sweeper: site scanThe new Sweeper module crawls your site from a start URL on a browser runner and reports console errors, uncaught exceptions, requests that returned 4xx/5xx, broken links, blank and slow pages. It only sends GET requests: no clicks, no form submissions. A pass/fail threshold fails the CI job. 50 pages without the module; up to 5,000 and schedules with it, with a 7-day trial.Shared headers and header suggestionsHeaders that are the same in every request, like Authorization or X-Tenant, are written once and go with the HTTP, GraphQL, WebSocket and SSE steps of every scenario; values may use variables. When the same header is written in several scenarios, the editor offers to move it to the shared area. Standard header names are suggested while you type (Aut → Authorization).New mock templatesReady mock templates for SendGrid, Twilio IVR, SMTP (over HTTP), Slack, Teams, the Turkish central bank's FX rates and the fictional ABC Bank virtual POS. None of them reaches a real service; no e-mail, SMS or payment is sent.Run from the list, a step name that looks editableA Run button next to Edit in the tests list opens the test with the run dialog open. The step name sits in a box with a pencil icon, and leading and trailing spaces are removed from URLs. Data sources in the menu is now Infrastructure.Search the menuThe search box at the top of the left menu stays in place while the menu scrolls. Type a few letters of a page's name; Enter opens the first match and Esc clears it.Durations as number + unit, errors at the fieldThe start delay and the load stage durations are entered as a number and a unit (ms, s, min, h). Validation errors are red: the invalid field gets a red border and the message sits under it. Scenario names may contain Turkish letters and hyphens.Quick install: a one-line commandIn Infrastructure → Add, kinds with a script install with a one-line command (bash for Linux/macOS, PowerShell for Windows). The script creates the read-only monitor user, shows each command first, asks before changes such as a restart, tests the connection and adds the source to Spitfire itself. The setup token is valid for 1 hour and single use; the admin password is stored nowhere. The Manual install tab keeps the copyable steps.Windows setup wizard: setup.exe and .msiSpitfire-Setup.exe installs with Next, Next and no terminal: it checks Docker Desktop, WSL 2 and the ports, streams the install output and adds Update and Uninstall to the Start menu. Spitfire.msi installs silently for Intune or GPO. No administrator rights needed. The files are not code-signed yet: in SmartScreen, More info → Run anyway.macOS installer wizard: Spitfire.dmgDrag the app in Spitfire.dmg into Applications and install with Continue; opened again, it updates, repairs or uninstalls. Since 1.4.14 it is signed and notarized by Apple: on first open just press Open when macOS asks.Kubernetes cluster viewNode, namespace, workload (Deployment, DaemonSet, StatefulSet), pod and container tables with use against requests, limits and allocatable. Findings for OOMKilled containers, CPU (throttling) and memory near their limits, missing DaemonSet pods, MemoryPressure and DiskPressure.All metrics: the metric explorerEvery metric a monitored system sends is on the All metrics tab, searchable and grouped by prefix; Prometheus endpoints too. Draw one as a rate or a value, filter or split it by attribute, and pin it to the dashboard with a threshold finding.Slow queries, locks and cache hitsTop queries (literals hidden) and sessions waiting on a lock on PostgreSQL, MySQL, SQL Server and Oracle, in one table; cache hit ratios on the dashboard. The queries come from pg_stat_statements, performance_schema and the system views, with the read-only user.Oracle, Redis, MongoDB, RabbitMQ, Memcached depthOracle's top SQL and tablespace use; Redis slowlog and replicas; MongoDB profiler, replica set, WiredTiger and current operations; RabbitMQ alarms and queues without consumers; the Memcached slab view, connection saturation and evictions.Kafka JMX: the broker dashboard and consumer lagThrough the JMX exporter: broker rates, under-replicated partitions, the active controller, heap and GC; consumer group lag in its own table. A Kafka cluster is one monitored system.Elasticsearch depth, OpenSearch (Beta)Shard states, pending tasks, JVM heap, free disk and watermarks, indexing and search rate and latency, thread pool rejections, circuit breakers; a largest indices table. The same view is open for OpenSearch as Beta.Cassandra (Beta)A Cassandra cluster is read through the JMX exporter: nodes, read/write latency, pending compactions, timeouts and unavailables, heap and GC. One cluster is one monitored system.Cloud (Beta): AWS, Azure, Google Cloud, vCenterMetrics of AWS, Azure and Google Cloud accounts and the hosts and VMs in VMware vCenter are read through the Tower with a read-only account. The setup script grants the read-only access and tests the connection.Docker (Beta): the container viewEvery container's CPU, memory and limit, network, block I/O and restarts on a Docker host, filtered by Compose project or image. Findings for containers restarting, near their limit, throttled or stopped.InfluxDB, ClickHouse and Couchbase (Beta)InfluxDB 2.x and 3.x, ClickHouse (slow queries included) and Couchbase join the monitored systems as Beta: a read-only user, a ready dashboard and threshold findings, like the other databases.Servers: a Linux, macOS and Windows agent; IIS (Beta)A native agent installed with one line (systemd, launchd, a Windows service) sends CPU, memory, swap, file systems, disks, network and services to the Tower; every server has its page under Servers. Findings for full file systems, load, stopped Windows services and failed systemd units. IIS app pools and the request queue through the Windows agent, as Beta.SNMP network devices (Beta)Switches, routers, firewalls and load balancers are read over SNMP, read-only (v3 first): interfaces, resources, sensors, BGP/OSPF and traps per device. Ready device profiles and read-only user commands for Cisco, Juniper, Arista, HPE Aruba, MikroTik, FortiGate, Palo Alto, Check Point, F5, pfSense/OPNsense and Ubiquiti.Java application servers (Beta)Tomcat, WildFly / JBoss EAP, WebLogic, WebSphere Liberty / Open Liberty and any JVM with JMX on (Spring Boot, for one): heap, GC, threads, request rates and errors, thread and connection pools. Traditional WebSphere is not supported.ZooKeeper, etcd and Consul (Beta)The leader, quorum and members, raft, WAL fsync p99, DB quota use; a critical finding for a cluster without a leader or quorum. One cluster is one monitored system.ActiveMQ Classic and Artemis (Beta)Over Jolokia: queue depth, enqueue/dequeue rates, DLQs, queues without consumers and broker memory; a queue table and threshold findings.Nginx and HAProxy depthThe Nginx JSON access log gives status classes, p95, upstream and peer tables with no module and no agent; NGINX Plus peer states too. HAProxy backend, server and frontend tables; a backend with no server UP is a critical finding.Infrastructure: database and service metricsPostgreSQL, MySQL, SQL Server, MongoDB, Redis, RabbitMQ, Kafka, Elasticsearch, Nginx and more connect with a read-only user; the managed Collector beside a Tower gathers their metrics. Oracle is open as Beta.A copied grant, a named failureCopy and run the read-only grant for each kind, then test the connection from the Tower. A failed test names the reason: DNS, refused, timeout, TLS, auth, permission or database not found.Ready dashboards and KubernetesEvery monitored system comes with a ready dashboard and simple threshold findings. Kubernetes clusters are added as cloud accounts/clusters, MongoDB Atlas and Prometheus endpoints as monitored systems.Monitored system and cloud limitsNext to servers, the license now has monitored system and cloud account/cluster limits: Free 1/—, Starter 2/1, Growth 5/2, Scale 15/5, Enterprise unlimited, with add-on packs when you need more.k6 scripts run as they areYour k6 script or k6 archive runs on runner-k6 without converting, and the results land in the Spitfire report. The script's thresholds are proposed as the test's own; k6's verdict shows as the "k6 thresholds" check.JMeter controllers and CSV filesRequests inside a JMeter plan's logic controllers are kept in order, with what each controller decided reported. CSV Data Sets become data bindings; the CSV files you pick with the .jmx are uploaded and bound.Postman test scripts are convertedTest scripts in a Postman collection become extractors and checks. The import page shows where each script was and how much of it was converted.Create a test from APMA service's real endpoint mix, the observed peak and p95 thresholds make 1.5x, 2x and breaking-point tests in one click. A Recommended tests panel and a Regressions card against last week or the previous version; regressions reach your channels too.Findings written to your monitoring toolTest findings can be written as events to Zabbix (trapper, acknowledge), Dynatrace and Instana, so the team sees what the test found in its own tool. Off by default; turned on in the connection settings.High availability and disaster recoveryWhat happens while the controller is down, scheduled encrypted backups (Docker and Helm), a restore drill, update and rollback downtime, in one guide. When a runner drops mid-run, the report and the findings say so.Spitfire 1.1: APM on your own serversYour services' traces, logs and metrics are no longer limited to a run window: the new APM menu keeps them continuously, in ClickHouse, on your own servers. Retention follows your plan: Free 1 day (1 GB a day), Starter 3, Growth 7, Scale 15, Enterprise 30 days; an admin can shorten it, not exceed it. Nothing is stored until an admin consents.The APM module in one click from the UINew installations come with the APM store on (--without-apm turns it off). On an installation without it, the button on the APM page installs it; no terminal needed.Services: requests, errors, avg/p95/p99Every service's requests per minute, error rate, average, p95 and p99 duration and status codes in one table. The service detail has its endpoints, exceptions and versions: a red arrow when a version is worse than the one before.Distributed traces and "Why did it break?"Find traces by service, endpoint, duration, status, attribute or trace id; in the waterfall, each span's attributes, events and logs. The Why did it break? panel ranks the likely causes with a score: the deepest failing span, the span that took the most time, a host under pressure, a new version. It is an estimate that shows where to look first.The Errors tab: every error in one placeYour services' errors in one list: search as you type, with filters that show their counts. Go from an error to its span and back in one click; the span detail opens with its attributes, events and logs.APM charts that filter as you clickOn trace search, duration over time (a point opens its trace) and a duration histogram (a bar narrows the duration filter to its range). On Errors, errors over time by type, following the filters; clicking a type adds it to the filter. Request and p95 trends on services, CPU and memory lines on hosts, a time ruler on the waterfall.Topology: who calls whomServices, databases and external systems on a map drawn from the spans. Every arrow shows requests per minute, p95 and error %; a failing call is red. The same is there as a table too.Host ↔ app mappingA host's page lists the apps running on it; the services list shows which hosts each app runs on. OS, runtime, container and k8s details; CPU and memory from the Tower.A guided setup wizard per languageThe Add a service wizard asks for the language, framework and version, creates the token itself and fills the numbered steps with your values, each with a copy button. At the end, the Did data arrive? step waits for the first span.OTel guides per framework and versionJava (Spring Boot, Quarkus, Tomcat, WildFly, WebLogic), .NET (ASP.NET Core, IIS, WCF), Python (Django, Flask, FastAPI, Celery), Node.js (Express, NestJS, Fastify, Next.js), PHP (Laravel, Symfony, Slim), Go (net/http, Gin, Echo, gRPC), Rust (axum, actix-web, tonic) and the new Ruby (Rails, Sinatra). Every guide is numbered steps ending with a check that prints OK.Browser and mobile: React, React Native, FlutterThe browser and app key (sfrum_) only writes traces and logs and cannot read; only from the origins you allow, within a per-minute limit. The React, React Native and Flutter guides link to your backend traces with traceparent; Flutter uses a community package, as there is no official Dart SDK.Spitfire 1.0: Load Testing & APMWith Spitfire 1.0 the product's name says what it does: Load Testing & APM. The load test finds how much your system carries; synthetic monitoring, Tower metrics and traces watch production around the clock. Spitfire has its own APM, and works with the APM you already have too. The new About Spitfire dialog in the user menu shows it with the version.Starter 1,000 VUs and 5,000 requests/s; a founding customer offerThe Starter plan's limit went from 500 VUs and 2,500 requests/s to 1,000 VUs and 5,000 requests/s. A Starter key bought earlier keeps the limits written in it; for the new limits ask support@usespitfire.com for a reissued key. The first 10 yearly Starter, Growth and Scale orders get 30% off the first year; no code needed, and the next 2 renewals cost the order day's list price. The places left are in the pricing section.Demo store and first steps: see a breaking point in two minutesInstalling with --with-demo (Windows: -WithDemo) adds a demo store next to Spitfire that runs on a deliberately small connection pool, with the ready test "Demo store: find the breaking point". The first steps list on the Overview page starts it as a breaking point run, opens the result and the finding, then leads to templates, import or a Chrome recording for your own system. Only local runners reach the demo app; no port is opened outside.Spitfire Recorder: record a flow in Chrome, get a testWith the Spitfire Recorder Chrome extension, use your site like a user (sign in, cart, checkout), press Stop and the recording goes straight to your own Spitfire. Correlations like CSRF, JWT and OAuth tokens, __VIEWSTATE and a new record's id are found and checked against the recorded response; the ones you accept become extractors and variables. Passwords, Authorization headers, card and Turkish ID numbers are masked before the recording leaves the browser, and cookies are not recorded. The extension has no analytics and sends nothing to Albiferro.Run JMeter, Gatling and Locust tests as they areA JMeter plan, Gatling simulation or locustfile now runs without converting it: the test runs on its own engine in that engine's runner image, and every sampler or request shows as a step in Spitfire; thresholds, comparisons and notifications work as usual. The load comes from Spitfire's scenario or from the file itself, with the license's VU and duration limits as the ceiling. Since these files can run code, engines are off by default; an admin turns them on after a warning, and the audit log records it.Deep APM: run markers and the slowest traces in Dynatrace and InstanaSpitfire has its own APM, and works with the APM you already have too. Dynatrace and Instana connections can write a marker to the APM at the start of a run, at each load stage and at the end (off by default). For every stage the services' p95, error rate, host CPU and memory and the 3 slowest traces are read; the APM card on the Backend tab and the report show them with links into the APM. If the APM can't be reached the run goes on, with a warning.Executive summary: a one-page PDFReport → Executive summary downloads a one-page PDF for a non-technical reader: the one-sentence result in a coloured box, Low/Medium/High risk (the colour always with the word), the weak point in plain words, three numbers, up to three pieces of advice and a comparison with the baseline or the previous run. With a target load, a breaking point below 80% of it is High risk. Share links open it too: …/r/<token>/summary.A run opens with "Result: …"A finished run's page and the HTML/PDF report open with a one-sentence verdict ("The system met every target at up to 500 VUs."), then 3-5 short lines: p95 against the threshold, the error rate, the load handled and the breaking point or the main findings. During a live run the progress line says what the load is doing now: ramping up 120 → 500 VUs, holding with the time left, ramping down.The feedback form: 30 days as a thank-youThe Feedback form in the user menu asks for a 0–10 recommend score, what you liked and what to improve; the What is sent box shows exactly what goes out. Send opens the usespitfire.com page with your answers filled in and you decide there; the Spitfire server sends nothing. The thank-you does not depend on the score: 30 days on a paid license, a 30-day Growth key on the free edition. A public review earns 15% off the next yearly renewal.Verify a release before installing: SBOM, vulnerability scan, cosignEvery release comes with SBOMs of the source and the images (SPDX and CycloneDX), a vulnerability report for each SBOM and cosign signatures of the files; the Docker Hub images are signed. A critical finding stops the release. Your security team can check the signature and the SBOM with its own tools before installing.Add a service: Go and Rust guides, OTLP/gRPC on :4317The OTel guides now cover seven languages: Go and Rust next to Java, .NET, Python, Node.js and PHP. In these two compiled languages you add a file and a few lines to the service and rebuild; for Go without code changes there is OTel eBPF too. The Add a service wizard on the APM and Towers pages asks for the language, environment and service name, fills the commands with your values and waits for the first span or metric in its Did data arrive? step. Next to OTLP/HTTP, Spitfire's receiver takes OTLP/gRPC on port 4317 with the same token.Connect a service (OTel): Java, .NET, Python, Node.js, PHP guidesThe wizard on the Towers page asks for the language and runtime version, the addresses, keys and service name; the language's guide opens with those values and the version's packages written in, ready to copy and paste. Every guide has Linux (systemd), Docker, Kubernetes and Windows steps and IS_OTEL_ENABLED=true/false to switch it on and off. The pre-release checks apply the guides as written with sample services: traces arrive in the run's window, metrics through a Tower.The System page and the guard with a server CPU limitThe new System page shows the Towers and the last day of every server, database and service in hourly summaries with a status (busy, no readings) on one page. The guard now watches the target server's CPU too: when the chosen servers cross the limit the load steps down, over the stop limit the run stops; the report names the server and its value. The test does not overload the server it measures.A Tower in one command, our own Collector image, more databasesThe New Tower dialog gives the install command for Linux (a systemd service), Windows (a Windows service), Docker and Helm with the token filled in; the install waits for the Tower to connect. install.sh docker --tower on and Helm bring a Tower and a Collector reading the host with the installation. The Spitfire Collector image carries only the components needed; the config wizard adds MySQL/MariaDB, SQL Server, Oracle, Redis, MongoDB and Prometheus endpoints next to PostgreSQL, with passwords kept out of the config.Alertmanager and Grafana alerts in the run report; syslog and SNMP countsTools sending Alertmanager, Grafana alerting or JSON alerts can post them to Spitfire or to a Tower with their own token; the alerts show on the events of the run reports they overlap, resolved ones close. Which alarm fired during the test is in the report. A Tower can optionally count syslog messages and SNMP traps per host and severity; messages are not kept, only counts.Report an error in one click: the ticket opens filled in on DesktikA page's server error box, the error notice and the error page have a Report button. It opens the Desktik support portal in a new tab, signed in with your usespitfire.com account, with the ticket filled in with the error message, its code, the page, the time, the Spitfire version, who reports it and the browser; only Send is left. Without internet a notice says so and the filled-in text can be copied to share from a connected device at app.desktik.com. Nothing is sent by itself.Wallboard: the team watches the live run on a TVThe Dashboard's Wallboard button opens a page with no menu that fills the screen; Full screen puts it on a TV or a projector. Up to 4 running runs update every second in their own panels: VUs, requests/s, p95, error rate and the last 5 minutes' charts. Beside them one tile per synthetic monitor, down ones first; at the top open incidents, 24-hour uptime, the 7-day pass rate and runners. The cursor hides when the mouse rests, the screen stays awake and the text grows enough to read on a 4K TV. For the TV, a user with the user role, which only sees results, is enough. While the mouse moves over one chart, every chart of the run draws a line at the same second: the p95 and errors at the moment the request rate dropped, at a glance.Alert rule: don't alert when only one location is downOn a synthetic monitor with several locations, pick "When most locations are bad" under Alerts and a round is bad only when more than half the locations failed or were slow; a tie is good. Example: of fra, ist and ams only fra lost its network, so no alert opens and the monitor shows a "partial problem: fra" badge. When two are bad, the monitor.down notice names the rule and says "2/3 locations bad: fra, ist". The monthly SLA's overall row is counted the same way, so the alert and the SLA agree. The default stays "any location"; existing monitors don't change.Frozen report: the same report a month laterWhen a run ends its report is built once and kept; HTML, PDF, the summary CSV, shared links and the Slack/e-mail/webhook notices all come from that record. Retention can delete the per-second data and the charts stay; the test's baseline can change later and that day's comparison verdict doesn't, the report just notes that the baseline changed since. The PDF you gave a client or an auditor shows the same numbers when downloaded today. The Report menu says when it was frozen; while the per-second data remains an admin can rebuild it, and the old record's digest goes to the audit log.Warm-up: keep the first minutes out of the resultSet Options → Warm-up (e.g. 1m) in the test editor and thresholds and performance budgets look only at the requests after it; the slow first requests while JIT, caches and connection pools fill no longer fail the test. The run page, the HTML/PDF report and the CSV give after warm-up and the whole run side by side. With ramping load models the new Stages table shows each stage's target and the measured VUs, requests/s, p95 and error rate, so you see at a glance at which step things started to slow down. Retention now trims per-second data mid-month too, run by run, and the run page says when it has.Pacing: each VU starts its iterations at a set intervalIn Constant VUs, Ramping VUs and the iteration models, Load model → Pacing starts each VU's iteration at a fixed interval (e.g. 30 s) or at random between min and max; when an iteration is quick the VU waits out the rest. That gives a real-user load like "one order a minute" and a predictable rate from a VU count: 100 VUs every 30 s → about 3.3 iterations/s, shown on the plan line. When an iteration takes longer than the interval the next starts at once and the pacing_missed counter goes up; missed starts are not made up in a burst, and above 0.1% the report shows a finding. A JMeter Constant Throughput Timer becomes pacing on import.Monthly SLA report: a PDF and CSV for your clientThe Monthly SLA card on the Synthetic monitoring page shows each monitor's SLA over the chosen calendar month (good checks / checks that ran), Met or Breached against its target, the outage minutes and the minutes the target allows: a 99.9% target allows 44.6 minutes in a 31-day month. A PDF with the workspace's branding and a CSV download in one click, with incidents, the worst day, locations and the month before. Checks skipped for maintenance aren't counted, and days are kept 400 days. Tick monitor.monthly on a channel and the report arrives by itself on the 1st at 09:00.Apdex: user satisfaction in one scoreEvery run gives one score between 0 and 1 for how satisfied users would be with the response times. A request finishing within the target T is satisfied, within 4T tolerating (half), slower or failed frustrated: 800 satisfied and 150 tolerating out of 1000 requests make (800 + 150/2) / 1000 = 0.875. The run page has an Apdex tile with its reading (excellent, good, fair, poor, unacceptable) and an Apdex column for each step; the report, the CSVs and comparisons have it too. T is in the test's options (500 ms by default) and can change per step; as a threshold, write apdex rate>0.9.VUs, iterations, runners, locations: what each field means and how it adds upIn the load model, the Run dialog and the breaking point form, the orange "i" next to the VU, iteration, iterations/s, runner and location fields explains the term with an example (like 10 iterations/s × 3 requests = 30 requests a second). The breaking point form shows its units and the requests per second. The new "How the math works" docs section shows every step with numbers, from VUs to requests/s, duration, splitting over runners and locations, and the license check.Tests from an API document sign themselves inImporting an OpenAPI / Swagger document finds the login operation (like /login or /auth/token) or the OAuth2 flow; type the username and password and try them once with Try sign-in. The test starts with a step that signs in once for each virtual user and takes the token, and every request after it sends that token. Even when the address can't be reached right then, the sign-in step is ready in the test; when nothing is found, give a token or API key yourself.So the firewall doesn't call it DoS and the APM tells test traffic apart: the traffic markEvery run request carries the X-Spitfire-Run header and spitfire.run in W3C Baggage, so APM and logs tell test traffic from real traffic. On the Runners page you can turn on a secret traffic key of your own: a firewall/WAF rule skips the rate limit when the runner's IP and this key come together. The docs have ready examples for Cloudflare WAF, nginx, AWS WAF and the OpenTelemetry Collector.Copy the runners' IP addresses in one clickThe Runners page lists the connected runners' public IP addresses with a copy button, and each runner's address is under its name (with a note when it's private). No need to look up the addresses for the firewall on the runner machines.The breaking point form shows its math and how to fit the limitThe capacity test's duration, top request rate and automatic VUs are written in the form with their formula (e.g. 20 steps × (60 s + 10 s ramp) = 23 min 20 s). When a value exceeds a license limit, it says which field to lower to which value, and clicking the suggestion sets it; a link next to it opens the plans that raise the limit. Automatic VUs no longer exceed the license's VU limit.Error messages say what to do, with clickable linksA license limit error also says what to lower to what; addresses in error messages are clickable. The run charts' "no APM connection" message links to the APM page. Guard's Custom opens its fields, and the test editor's button says "Save" for an existing test.A runner reserved for synthetic monitors on the free editionOn the free edition with two or more runners connected, one only runs the synthetic monitors' checks: the Runners page marks it "Reserved for synthetic monitors", the run dialog does not let you pick it, and the controller refuses a test on it even through the API. Monitors keep checking during a test, and a test never waits for a check. A paid license opens every runner to tests.Template descriptions say the values you enterSet a built-in template's limit to 200 ms instead of 800 ms and its description says 200 ms; concurrent users, duration and request rate too. The new test's description is saved with the values you entered.SCIM token revoke, revoked tokens out of the wayThe SCIM token can now be revoked without making a new one. Revoked API, Gate and runner tokens and Towers leave their lists; "Show revoked" brings them back. Requiring two-step verification lists the accounts that have not set it up yet and warns first when the admin's own account is one of them. After an OIDC issuer change, "Test" no longer asks for the secret again.28 new flows in the pre-release checksThe pre-release checks grew: the run report and stopping a run, templates, OpenAPI/traffic/file import, fake data, SSH tunnel, integrations, Towers, runners, SSO, two-step verification, the audit log, the license, help pages and every "i" explanation are tested end to end in Turkish and English. What they found is fixed too: the Towers page that failed to open while a Tower had not sent metrics yet, the audit log that did not refresh after a CSV export, and Collector settings errors shown in English.Every release, after the whole interface is exercisedBefore a release is published, every button, menu, switch and select on every page is now clicked; each dialog's form is sent empty to check its validation, each link is opened and each chart is checked to be drawn. Sign-in, passwords, API tokens, connection tests, users, groups, schedules, monitors, mock services, data files, the test editor and starting/stopping runs are tested end to end. If any of these or the unit tests fail, the release does not ship.The mobile app on the sign-in pageThe sign-in page now shows the Spitfire mobile app: a Google Play link and, on wide screens, a QR code to scan with your phone; the App Store version is coming soon. Installing, adding a server and turning notifications on are covered in the help.Tower and Collector: servers and databases during the testA Tower is a small gateway in your network: it takes server and PostgreSQL metrics from Collectors, keeps them on disk while the controller can't be reached and forwards them over one outgoing link; nothing connects into your network. On the Towers & metrics page, the Collector config takes a few clicks (the token and the password stay out of it). On the run report's Backend tab, CPU, memory, disk, network and database connections, commits and deadlocks, with no connection added to the test. Retention follows your plan (7 days on the free edition up to 365 on Enterprise), storage consent comes first, and the free key includes 1 server + 1 database.Templates: filled with a suggestion, yours to changeIn a built-in template the p95 response time limit, error rate, concurrent users, request rate and duration come with a suggested value; write 500 ms or 1 s instead of 800 ms, or go back with "Back to suggestion". Required fields carry a red asterisk, and the test isn't created until they're filled. My templates: save the form's settings or an existing test as your own workspace template. Nine new built-in templates: login + token, spike, soak, GraphQL, WebSocket, gRPC, Redis, MongoDB and RabbitMQ.Guard: a test that doesn't wear the system downWhen a test's own p95 or error rate crosses your limit, Spitfire steps the load down with a live req/s ceiling, climbs back slowly once things calm down, and stops the run at the stop limit. Every step shows in the event feed, on the run page and in the report. In the editor under "Thresholds and guard", with presets; off by default.One-click update (Docker)./install.sh docker --updater on turns it on, and the License → Version and updates card shows an Update now button. The database is backed up first, running tests are waited for, and only a release from the signed list is installed. Only a separate small updater container reaches the Docker socket; the controller never does. If something goes wrong, the UI shows the error, the backup's path and the rollback command. Off by default.Health data sharing, only if you turn it onSpitfire sends nothing unless an installation admin turns it on. Then only an error/health summary goes out once a day: version, OS, counts of Spitfire's own fixed error messages, crash function names, runner counts. Never test content, URLs, hostnames, IPs, users, e-mails or the license; the card shows the exact payload. SPITFIRE_TELEMETRY=off disables it entirely.Every system with its own icon, a collapsing menu, thresholds in plain wordsEvery system on Integrations, APM and Connections shows its own icon (Datadog, Prometheus, PostgreSQL, Kafka, MQTT…); on an empty Connections page, click a type and New connection opens with it. The left menu collapses to icons with « and opens again with ». In the editor, thresholds read in plain words (e.g. 95% of measurements must stay below 500 ms), and raising Start VUs lifts the lower stages. The license key box follows the state; the new-release check runs every hour, and updating over an installed release goes through.Every system you can connect, at a glanceThe Integrations page opens with a card showing everything Spitfire can talk to, in three groups: notification channels (Webhook, Slack, Teams, e-mail, SMS), issue tracking (Jira, ServiceNow, Azure DevOps, GitLab, GitHub, YouTrack, Redmine, Linear) and APM. The ones your license lacks are faded and locked, with the plans they come with and an Upgrade the plan button. On the APM page, open standards (OTLP, Prometheus, Tempo, Jaeger, Loki) and monitoring events (Zabbix, Datadog, New Relic, Dynatrace, Instana) sit on separate rows. Before the first key, the License page shows only the key box, the installation ID and where to get a key.Templates in plain words, no JSONPathTemplate cards say what they do and what you will know at the end. In the list → detail and sale-day templates you type the name of the field that fetches the detail instead of a JSONPath (id, uid, name…); Spitfire finds the first record wherever the list sits in the response. With no tests yet, the Compare page shows the way with a Create test button. A green Enter license key button on the free edition, Upgrade the plan on the License page, and data retention fixes a day count above your license's limit by itself.A clearer License pageThe key box sits right above the limits, in an orange frame. An accepted key shows a green "installed" and the page reloads by itself; a refused one shows a plain reason and what to do in red. The Data retention card starts with a "What is kept right now?" summary: how many recent days are kept, older data deleted day by day each night, what is never deleted. The License pill in the top bar brings you back in one click.Stronger passwords with a live checklistNew passwords need at least 8 characters with an upper-case letter, a lower-case letter, a digit and a special character; letters of any alphabet count. On the first-admin, add user, reset and change password screens, the list under the password box ticks each rule green as you type and shows the missing ones in red. The server enforces it too; existing passwords and LDAP/SSO accounts are not affected.The UI opens in the installation's languageInstall with SPITFIRE_LANG=tr and the UI opens in Turkish, even if your browser is English; with en, in English. A user who switches the language in the top menu keeps their choice. An update without a language keeps the earlier installation's. On Helm, controller.uiLang picks it.Cleaner screens, install notes saved to a fileLong explanations no longer sit on screen: click the orange i next to the heading to open them. An i you have not read pulses gently and stops once opened; warnings and errors stay visible. The commands on the first setup screen have a copy button. When the install finishes, it offers to save the important lines it printed (address, setup code, uninstall command) to a log file; leave the path empty and it saves into the Spitfire folder and tells you where.A notification bell and a Rewards pageThe bell in the top bar shows unread notifications in a red badge. The welcome, your-rights and new-release notices are made by the installation itself and need no internet. The new Rewards page walks you through it: share on LinkedIn with a ready text for a 10% coupon, invite friends for up to 30% off. spitfire.tr announcements are opt-in: nothing is read until an admin says yes, and then the signed list is only read and nothing is sent.Eight issue trackers like Jira, with a version choiceAzure DevOps Boards, GitLab, GitHub, YouTrack, Redmine and Linear join Jira and ServiceNow: a problem opens one issue, and its repeats and recovery are comments on it. Setting up a channel asks which version you run (e.g. Jira Cloud or Data Center 9.x, Azure DevOps Server 2019); the connection test reads the server's version and tells you when it is older than supported or differs from your pick. The Zabbix connection got a version choice and user/password sign-in too (5.0 LTS and later).The install ends in a green boxWhen the install succeeds, a green box says “Spitfire was installed successfully.” with a clickable link to the UI. If it stops, it prints an error code starting with SF- and the support address; just e-mail the code and the output. The requirements page now lists the supported operating systems and their versions too.A bulleted summary at the end of the installWhat the installer prints when it finishes is now a bulleted list, with bold coloured labels and blue, clickable links. The license step sits right under the setup code. The network address says what it is for: colleagues on the same network reach the UI there. The local runners and the address remote runners connect to are separate items.The installer says what it will installOn Linux without Docker the installer asks first and only on yes installs and starts Docker Engine with the official get.docker.com script; the default answer is no. It does not install Docker Desktop on macOS or Windows (it links to it) and never installs Kubernetes. Before installing it reads the machine's vCPU, RAM and free disk, says which hardware profile it fits and lists the components, ports, rough disk and folder under What gets installed.Incident history and MTTRThe new Incident history card on a monitor's page lists the incidents of the last 30 days: when each started and ended, how long it lasted, outage or slowdown. On top sit the incident count, the total incident time and the mean time to recover (MTTR). The public status page for your customers now shows the recent incidents too; error messages and addresses are still not published.A status badge for your READMEAdd /badge.svg to a status page link for a small badge to put in a README or on your web page: "status | operational", or with ?show=uptime "uptime | 99.95% (30d)". Its colour follows the state, it speaks Turkish or English and refreshes every 30 seconds. The Status pages card previews both badges and copies their Markdown in one click.A status page for your customersPick monitors on the card of the Synthetic monitors page and press Publish: a link that needs no sign-in shows the overall state, each monitor's state, the uptime of the last 30 days and coloured bars day by day. Turkish or English, light and dark, refreshed every minute. The test name, environment and error messages are not published; New link retires the old one at once, and the JSON form works on your own site too.Synthetic monitors in Prometheus and GrafanaSpitfire's /metrics now carries the monitors too: up or down, the last check, the error budget left and its burn rate, the SLI, drift and planned maintenance. The Grafana dashboard in the examples repository has a new Synthetic monitors row, and teams on Alertmanager get five ready alerting rules, including a watchdog for a monitor that stopped checking. A user's token shows only their own monitors.A weekly watch summary every MondayOn Monday at 09:00 your Slack, Teams, e-mail or webhook channel gets last week's synthetic monitor summary: uptime and the week before's, bad checks, SLO status, the monitors drifting, the week's deploys and, worst first, each monitor's p95 change. Tick the new monitor.weekly event on the channel; to see it without waiting for Monday, press Send the weekly summary now.