14 changes since 14.0.104: 6 new or improved features, 1 bug fix and 7 maintenance changes.

✨ New and improved

• Cloud
  • Right-sizing from OneUptime's own readings, unusual-day alerts on budgets, budgets over MCP and CLI (#5300 part 3) (#5752)
  • #5702's follow-ups - health events and a case's cloud changes in the reader's words, App Platforms lists by cloud account, the hero's colon per language, the environment guide in every language (#6149)
• Cloud Streaming: The command checks AWS's log subscription limits before it deploys, Step Functions logs land on their state machine, and stream keys and refused deliveries hardened (#5713 follow-ups) (#6137)
• Status page logos travel inside their emails, whatever they were uploaded as (#6135)
• Real-time rules that count and correlations; playbooks get a Stop step and steps that wait on several results (#6138)
• Cloud AI agent security: a key of each agent's own, per-change Google Cloud tokens, older key versions destroyed, refused agents on the feed, and Enable AI stops a held command (#6148)

🐛 Bug fixes

• Forms: Every field's error in one red that can be read, no Clear on a required pick, and Create Monitor's step cards as sections at every width (#6145)

⬆️ Upgrade

Back up first, then follow the upgrade guide. On Docker Compose, run npm run update on the release branch. On Kubernetes, helm upgrade to version 14.0.105 of the chart.

Full changelog: 14.0.104...14.0.105

Contributors: @kvncrw, @simlarsen

373 changes since 14.0.103: 137 new or improved features, 103 bug fixes, 6 performance improvements, 27 documentation updates and 100 maintenance changes. The busiest areas: Openstack, Data Observability and Workflows.

✨ New and improved

• Openstack
  • Monitor an OpenStack cloud: inventory, agent, pages and alerts (#5789)
  • Compute, network and storage pages, and the guards that know OpenStack (#5798)
  • Connect a cloud with no agent, guides per deployment tool, control plane, real VM metrics, logs, events, audit trails and traces (#5808)
  • An investigation of a region's API says what the region takes with it, by project (#5818)
  • A probe says each instance's address and which pool holds each volume (#5819)
  • The agent reads load balancers, bare metal, stacks, clusters, shares, DNS, object storage, databases and Gnocchi where a cloud runs them (#5820)
  • How many more of each flavor fit, worked out host by host with what Placement would decide (#5822)
  • The Services page says when an operator forced a Nova service down (#5826)
  • An instance's page and list say which image it booted from (#5829)
  • Real CPU, memory, disk and network per instance and hypervisor, read true from libvirt or Ceilometer (#5831)
  • A DHCP agent's page lists the networks it serves, and a network names its agents (#5832)
  • Traces read as OSProfiler really writes them, and say what a stock cloud leaves out (#5833)
  • A case names the project and instance an audit event touched, Who changed this finds them, and every tool's guide says how to turn the audit trail on (#5835)
  • An alert template for an image stuck on its way, which recovers when it moves (#5837)
  • A pool's page lists its volumes, where a probe says which they are (#5838)
  • A cloud joins up with hosts, Kubernetes nodes, services and Ceph, and never on a guess (#5840)
  • A cloud of any size says what it costs, and its pages and rows hold up at 200,000 instances (#5842)
  • Connect a cloud with a probe from the dashboard alone, and see exactly what failed (#5846)
  • The region you pick narrows Topology, Metrics, Logs, Traces, Incidents and Alerts, and each project's quotas are read once an hour (#5848)
  • Traces say what each release really does, and warn off the OpenTelemetry version that hangs Neutron (#5849)
  • A Capacity page says how many more fit, and when things fill up (#5852)
  • Bare metal nodes reach a cloud's inventory with health rules and alerts, and a node no longer counts as a hypervisor (#5857)
  • A rule on a project or domain is a real condition, and the map joins instances to their networks from what a probe says (#5858)
  • OneUptime AI investigates a cloud, read-only (#5859)
  • Flavor counts match a hand calculation on three real hosts, fragmented estate included (#5862)
  • Alerts on what a cloud stands on, and one failure opens one incident (#5863)
  • Heat stacks and Magnum clusters appear on a cloud that runs them, and leave no trace on one that doesn't (#5865)
  • A probe checks the cloud from outside, and can boot an instance in it (#5872)
  • A Magnum cluster with the Kubernetes agent links both ways with its Kubernetes cluster (#5874)
  • Manila's shares appear on a cloud that runs it, with pages, alerts and Topology, and leave no trace on one that doesn't (#5878)
  • The AI agent can make the ten changes a person approves, and refuses everything else (#5882)
  • A probe reads Manila: every share, its backend and access rules, and the services only it can see (#5883)
  • A cloud that runs Ironic gets a Bare Metal section, with its nodes and alerts, and a probe reads them (#5885)
  • The checks from outside, recorded on a real cloud (#5886)
  • Usage per project from inventory, your own prices, budgets, and what could be reclaimed (#5887)
  • Kolla-Ansible's real recording is in, and the guides say what ran on real deployments (#5891)
  • A stack's page lists the instances and volumes it owns, and each of those names the stack (#5893)
  • A cloud's Spend page says who's using what, at your prices or CloudKitty's, and Recommendations says what could be reclaimed (#5894)
  • The Capacity page adds up bare metal by resource class, and a probe can alert on a stopped conductor (#5897)
  • Trove's databases appear on a cloud that runs it, and a credential it refuses reads as that rather than as Trove being down (#5898)
  • A probe reads Trove too, and says why a database failed, which the exporter can't (#5900)
  • A product page for OpenStack, with its place in the navigation, the air-gapped page and the compare pages (#5901)
  • Designate's zones appear on a cloud that runs it, and what a stopped Designate can't say, it doesn't claim to (#5906)
  • OneUptime AI can fix an OpenStack cloud once you let it twice, and an evacuation always asks a person (#5909)
  • Gnocchi's backlog and processors on a cloud that runs it, with a probe's read of them too (#5911)
  • A probe reads Designate too, and can say what a stopped one can't say for itself (#5913)
  • A probe reads what CloudKitty rated, and a cloud's Spend page prefers it (#5915)
  • A workflow can propose a fix from an OpenStack alert, and a person approves it (#5919)
  • Swift's containers appear on a cloud that runs it, as far as a credential can read them, with a growth forecast (#5920)
  • A probe reads Barbican where a cloud runs it, and how many secrets and orders it holds, without reading one (#5922)
  • OneUptime AI reads Manila, Designate, Swift, Trove and Gnocchi where a cloud runs them, and still reads no secret (#5923)
  • A cloud running Octavia gets its load balancers, pools and amphorae, and alerts for the ones in trouble (#5925)
  • A probe reads Swift's storage servers where you name them: whether each answers, its replication and its rings (#5926)
  • One step makes a status page per region, and a failure or planned work tells only the projects it reaches (#5927)
  • A cloud's Status Pages page, and each project's notices and private page (#5929)
  • A probe-collected cloud shows Octavia's listeners and members, and charts a load balancer's traffic (#5939)
  • OpenStack-Ansible's real recording is in, and the guides say what Cinder needs for a credential limited by access rules (#5948)
  • A real Alertmanager's webhook from a real Kolla-Ansible, and Cinder read on both tools (#5952)
  • Trace every API request with nothing signed on 2026.2, through OpenStack's own tracing (#6011)
• Data Observability
  • Watch every table of connected databases for row count, size, last change and schema changes (#5647)
  • The Data Observability product in the dashboard - overview, tables, schema changes, databases and a page per table (#5654)
  • Recommended alerts for a database's tables, and a page per database (#5662)
  • The services that read and write each table, from their database spans (#5667)
  • Checks on a table's values - empty, repeated, invalid, orphaned, custom SQL - segmented, with alerts (#5669)
  • Suggested checks from a table's columns, accepted in one click and dismissed for good (#5670)
  • Pipelines and lineage from OpenLineage events (#5671)
  • Alerts on failed, slow and late pipelines, on a page per pipeline (#5673)
  • Dbt artifacts - models on their tables, lineage, tests as checks, builds as runs (#5677)
  • One issue per cause - a data alert whose cause is open upstream starts acknowledged (#5684)
  • Unused tables - each table's last read, from the engines' own read counters (#5687)
  • Amazon Redshift - tables, sizes, last writes and reads from Redshift's own catalog (#5690)
  • Snowflake - over its SQL API with key-pair authentication, the catalog read with SHOW at no credit cost (#5691)
  • Google BigQuery, with each Google Cloud connection's own service account (#5692)
  • Lineage from every view's definition and from BigQuery's and Redshift's query history (#5695)
  • Fivetran and Airbyte syncs as pipeline runs (#5697)
  • Data products, with a freshness SLA, status pages, an SLO and a badge (#5698)
  • Reconciliations between tables and systems (#5699)
  • OneUptime AI knows the data - table freshness everywhere, data alerts investigated from the data, checks written from a sentence (#5700)
  • ClickHouse and Databricks - table health, checks and lineage (#5701)
  • Production Review names what a pull request's migrations and dbt models break (#5703)
  • Dashboards of BI tools in lineage - a stale table's alert lists the dashboards built on it (#5707)
  • Warehouse spend - credits and bytes billed by job, team and table, failing jobs and unused tables (#5708)
  • Amazon Athena and Iceberg through the AWS Glue Data Catalog (#5710)
  • Oracle and MongoDB (#5711)
  • Finish the spec (#5784)
  • Translate what still read English (#5787)
  • Bill watched tables, and let a database be stopped (#5792)
  • Let OneUptime Cloud keep failing rows too (#5800)
  • Connect a Database walks two steps, the database and then how OneUptime reaches it (fffe35e)
• Workflows
  • Datadog and New Relic templates - each alert opens an incident or alert, pages on-call and resolves on recovery (#5683)
  • Pick linked records by name instead of pasting their IDs (#5949)
  • Custom fields are fields of the record in every step's settings (#5953)
  • Run a workflow again with the values an earlier run started with (#5956)
  • Describe a workflow to Ask AI, and it builds it or changes it (#5962)
• On-Call
  • Skip until…, Get cover and mobile cover are locked, with the reason, for members whose override permission covers only some policies (#5661)
  • Skip until…, Get cover and the mobile app's cover offer the on-call policies a member may cover (#5672)
  • Escalate a page to a level you pick, hand-off rows that fit a phone, plain-case on-call labels, and Execute On-Call Duty Poli...

54 changes since 14.0.102: 18 new or improved features, 9 bug fixes, 2 performance improvements, 3 documentation updates and 22 maintenance changes. The busiest areas: Docs, Network and Data Security.

✨ New and improved

• Network
  • Project admins choose the layout every network map opens on (#5601)
  • Add Meraki's devices as network devices, and the Inventory refresh follow-ups (#5392) (#5620)
• AI remediation: Security alerts linked to what they are about, Fixes turned on decide again, resolved records, and a lighter fixes block (#5589)
• Bulk actions: Select every matching exception, a table's Bulk Actions in the band over its cards on a phone, whole Health pills at 1280 (#5587)
• CLI: Every API resource as a command, /install/cli, a Developer > CLI page, and CLI docs for scripts and AI agents (#5581)
• Data Security: A store's Overview draws who can reach it and where its data can go on one map (#5644)
• Docker, Swarm, Podman, Ceph, Storage Arrays: What needs attention first by each product's rules, a page for every part, and two Swarm alerts (#5604)
• Logs: Saved views stay on the page they were saved on, the Logs page or one resource's Logs tab (#5593)
• Mobile App: Tabletop Home, Settings and sheets, the divider from a keyboard, Acknowledge again, Acknowledge and Escalate under a page's push, hunt pushes open their page, and the posture module's iOS side (not linked yet) (#5596)
• Security: Security Events and Data Security have a Security category of their own in the products menu (#5592)
• Slack and Teams: Switch project on the Teams Home, Ask in another project, durable Home choice, and confirmations that name the record (#5595)
• Security Events cases, next: case custom fields, note templates and numbering; AI chat tables' dates and titles; no plain-object lookups (#5577)
• A view-only workflow step keeps what only reads usable (code box Copy, More fields) and draws no action it cannot take; WorkflowViewOnlyGuard holds every Builder control to the view (#5580)
• Count only Runners that take AI code fixes, and say on the Runners page that global Runners do them (#5302 follow-ups) (#5583)
• Runner and App images at half the download, each started by a test; SSO line, setup card reverse case, host PostgreSQL port, server translator (#5584)
• CI and tests, next: release e2e on per-arch tags with a digest check, Dashboard-like fixture routers, settled page and box reads, OldestKeyMap, an offline group budget (#5586)
• Slack and Teams Declare Incident pick the affected resources, and link the alert of its channel (#5588)
• i18n(dashboard): the workspace Home choice model's words in every language (49c119d)

🐛 Bug fixes

• Docs
  • Every docs page starts with a doctype, so browsers draw it in standards mode (#5610)
  • The built stylesheet's address is read without a pattern made from text, and its test reads markup through the shared walk (ce7cd2a)
• API Reference: The reference links no web app manifest, so its pages log no errors (#5611)
• Dashboards: Every widget drawn once, at its own size, from the board's first frame; the dashboard opens on its loader (#5590)
• Install: The /install 404 lists the CLI's installer without the product's name (25edf9d)
• Network: The map's Make default and Clear default buttons take the shared busy look (1840a6c)
• Permissions: Each project's permission snapshot is kept under its own names, so two projects in two tabs never swap roles (#5306 follow-up) (#5585)
• Security: A case custom field's Map Value From lists its allowed values, as every definition table's does (f8b45ac)
• Website: The VMware page lists Host Not Responding and says twenty-five monitors; the queues mockup reads #5553's labels (a6b3dac)

⚡ Performance

• Docs: The docs and the API reference link a built stylesheet instead of generating CSS in the browser (#5618)
• Status Pages: Sign-in and status pages download only the reader's language (#5616)

📚 Documentation

• Data Security: A store's paths on one map, in all 17 languages (26adb8c)
• Developer: Every page that lists a Developer section names its CLI page, in every language (d5be24c)
• Pr Assets: The UI review's shots of Affected Resources on Slack's and Teams' incident forms (2d34e72)

⬆️ Upgrade

Back up first, then follow the upgrade guide. On Docker Compose, run npm run update on the release branch. On Kubernetes, helm upgrade to version 14.0.103 of the chart.

Full changelog: 14.0.102...14.0.103

Contributors: @simlarsen

55 changes since 14.0.101: 18 new or improved features, 16 bug fixes, 1 performance improvement, 3 documentation updates and 17 maintenance changes. The busiest areas: MCP, API and Telemetry.

✨ New and improved

• MCP
  • The coding agent plugin package for Cursor's Marketplace and the Claude directory (#5530)
  • Each release publishes the coding agent plugin to its own repository (#5540)
  • The ChatGPT plugin carries the coding agent skills, for ChatGPT and Codex (#5573)
• Analytics: The marketing webhook reports a project's first payment and a subscription ending, and a converted trial reaches Google Analytics (#5539)
• Blog: Posts link related posts, show their last update, and are sent to IndexNow (#5541)
• Hosts: Which hosts need attention first, each host's health and readings, four new alerts and uptime (#5535)
• Product marks: One-colour inline SVG everywhere, no logo images (#5521)
• Resource pages: Details cards say only what the hero and header do not, one casing in every language; a Proxmox part says when it cannot be read (#5553)
• Security: In your logs and traces under Connected Signals; monitors, cloud resources, serverless functions and a member's on-call and teams; the Security card in browser fixtures (#5522)
• Session Length: Admin Dashboard > Settings > Session Length - how long a sign-in lasts, for password and SSO sign-ins (Enterprise) (#5570)
• Sign In: Continue with Google on the sign-up and sign-in pages (#5542)
• Signup: Keep the Google Ads campaign a visitor came from (#5579)
• Telemetry: Leave a facet value out (is not) in the Logs, Traces, Exceptions and Metrics explorers (#5574)
• Terraform: People by email address in every resource, oneuptime_user by email, team members invited by address (#5536)
• VMware: World-class vCenter pages - Needs attention, each host's real state, a page for every part, Host Not Responding (#5575)
• Navigation, next: every side menu lights a row's list, only long words hyphenate (tested in WebKit), every trail title in every language (#5524)
• Terraform/OpenTofu provider audit (block 48): choices checked at plan time, imports that settle, secrets kept secret, rows imported by what they join (#5576)
• i18n(dashboard): the session length and Google Ads column titles in every language (4add1ba)

🐛 Bug fixes

• API
  • The OpenAPI spec lists the fields and operators a telemetry query takes, time ranges and attribute filters included (#5537)
  • The OpenAPI spec lists every field a list can filter, sort and select by, with the operators each takes (#5543)
  • The OpenAPI spec lists the fields a telemetry record is created with, and that an update changes none (#5552)
  • A telemetry create's spec lists the fields with a default too, and an update of an entity key is refused (#5554)
  • Telemetry records can be created through the API, and are kept for their retention (#5556)
• MCP
  • The website and README say how to connect an AI agent: the address, signing in and the plan it needs (#5527)
  • A filter written another way, such as MongoDB's $regex, is refused with an error that says how to write it, instead of matching nothing (#5533)
• Telemetry
  • A group by is held to the same column permissions as a select (#5551)
  • A create writes each value as its column's type, and refuses one that is not (#5555)
• Accounts: An invitation's sign-up link no longer sends the invitee's email and claim token to Google Analytics (#5532)
• Blog: AI answers cite a post's page, not its Markdown copy (#5531)
• Hosts: A failed systemd unit's alert names the unit and says how to look at it (eba396d)
• Permissions: A member limited to some labels or to what they own creates telemetry only for those resources (#5567)
• Signup: Long ad links and phone numbers typed with spaces no longer block sign-up, and Microsoft sign-ups keep their campaign (#5528)
• Website: The VMware page lists Host Not Responding and says twenty-five monitors; the queues mockup reads #5553's labels (689f603)
• Workspace: A reaction's name loses its colons in one pass (be5990c)

⚡ Performance

• MCP: AI agents load the tool list in about 59,000 tokens instead of 300,000, and look up a resource's fields with oneuptime_describe_resource (#5529)

📚 Documentation

• Sign In
  • Continue with Google's Google Cloud project and OAuth clients are set up (#5544)
  • Continue with Google's branding is verified and shown on Google's sign-in screen (#5545)
• Telemetry: The Services list's Technology column, as each language's Dashboard draws it (049ba46)

⬆️ Upgrade

Back up first, then follow the upgrade guide. On Docker Compose, run npm run update on the release branch. On Kubernetes, helm upgrade to version 14.0.102 of the chart.

Full changelog: 14.0.101...14.0.102

Contributors: @simlarsen

11 changes since 14.0.100: 4 new or improved features, 2 bug fixes, 1 documentation update and 4 maintenance changes.

✨ New and improved

• AI Sre: Approve OneUptime AI's fixes from Slack and Teams, each change's track record, what an investigation ruled out, and cloud changes in plain words (#5519)
• Kubernetes: What needs attention first, a page for every workload, pod and node, one set of health rules, and three new alerts (#5518)
• Monitors: Unusual-value alerts on a filtered or grouped metric are judged by that series' own history (#5525)
• Network: Compare WAN uplinks with loss and latency, and see what each Meraki network talks to (#5517)

🐛 Bug fixes

• Common: A remediation's track record and a cloud change in words stay in the npm package (efcb827)
• Website: The air-gapped Kubernetes page counts twenty-one alert templates, as the docs do (e970427)

📚 Documentation

• Kubernetes: The Networking menu's Routes, as each language's Dashboard draws it (92ffe04)

⬆️ Upgrade

Back up first, then follow the upgrade guide. On Docker Compose, run npm run update on the release branch. On Kubernetes, helm upgrade to version 14.0.101 of the chart.

Full changelog: 14.0.100...14.0.101

Contributors: @simlarsen

148 changes since 14.0.99: 50 new or improved features, 45 bug fixes, 1 performance improvement, 4 documentation updates and 48 maintenance changes. The busiest areas: Exceptions, Kubernetes and Monitors.

✨ New and improved

• Exceptions
  • Similar exceptions on the Overview, and OneUptime AI says which are the same bug (#5431)
  • Alert on new, regressed and spiking exceptions from an Exceptions monitor (#5440)
  • Affected users and a trend chart on every row of the exceptions list, from one stats call per page (#5443)
  • Merge a likely duplicate into an exception and unmerge it - the merged fingerprint becomes an alias at ingest, counts and occurrences follow (#5445)
  • Group new exceptions by the functions and modules of their own code, while existing ones keep their first fingerprint (#5447)
  • An activity timeline per exception - every triage, assignment, merge and regression recorded on the server, shown on the Overview (#5450)
  • Notify the person, or the members of the team, an exception is assigned to - the owner-added pattern, through each person's notification settings (#5451)
  • Saved views of the exceptions list (#5454)
  • Alert on exception spikes against what is normal for the hour of the week (#5506)
• Kubernetes
  • A cluster's Platform shows OpenShift, EKS, GKE or AKS, read from its telemetry (#5481)
  • The setup guide installs on OpenShift, and the agent docs say how (#5483)
  • The openshift preset pulls OpenShift's operators, version, Routes and DeploymentConfigs into inventory (#5484)
  • The openshift preset pulls the Services Routes point at, and names a pod's DeploymentConfig (#5487)
  • Routes, Services and DeploymentConfigs have pages of their own (#5489)
  • An OpenShift cluster's overview leads with its operators and update (#5491)
  • An OpenShift cluster is offered alerts on its operators and update (#5492)
• Kubernetes Agent
  • Restart and Job failure increases carry the workload names their counts carry (#5469)
  • An openshift preset, so the agent installs and runs on OpenShift (#5480)
  • An openshift-restricted preset, and OpenShift's control plane through service discovery (#5494)
• App Platforms
  • Split the app and container platforms out of Cloud into App Platforms (#5488)
  • Every app and its deploys - All Apps, an app's page, revisions, rollbacks, per-app requests, errors and latency (#5511)
• Inventory
  • Read what a device reports about itself, and tie an MX to its Meraki record (#5384) (#5495)
  • A discovery scan reads a host's MAC, serial number and firmware (#5384) (#5502)
• UI polish
  • Forms on a custom domain - refusals under their field, Edit without a fold, DNS Setup first, addresses left to right (#5478)
  • Agent install commands read whole where they are shown (#5455 review) (#5479)
• Website
  • An OpenShift monitoring page leads into the Kubernetes product (#5493)
  • An air-gapped page for each product (#5503)
• Dashboards: Restart charts show restarts per interval, falling back to the running count for older agents (#5467)
• Data Security: Who reaches the data, OneUptime AI for unplaced fields, six more one-click fixes and a case's data at risk (#5500)
• Forms: A form on its own custom domain, as status pages have (#5463)
• License: Manage the license on Settings > License in the Admin Dashboard (#5505)
• Monitors: One simple setup for every infrastructure monitor - what to watch with whether it sends data, the default alert it starts on, and the rest under More fields (#5432)
• Network: Meraki places clients on their switch ports and reports the cables the map was missing (#5497)
• Proxmox: What needs attention first, a page for each node, guest and storage, backup results and five new alerts (#5490)
• Rum: Follow one user or session through User Flows, page by page (#5514)
• Security: Playbooks that branch on a step's result, Add tasks, Entra ID and Google Workspace containment, and real-time detection rules (#5424)
• Traces: The traces list, Insights and the older span panel in all 17 languages (#5508)
• Translations: The translation backlog's next pass - form fields' help, Dismiss, the Products menu, Korean, Ceph and more in every language (#5512)
• Workflows: Every template in five steps or fewer, and who to page asked when it is created (#5430)
• File and Multiple files custom fields, and Forms file questions that fill them (#5436)
• Monitors follow-ups (#5281): grouped infrastructure reads in ClickHouse, SNMP walk Download, Monitor web UI, phone fixes, readable secondary links, overall status in words (#5437)
• Dashboards fit their screen and stack on phones; public dashboards get a dark theme logo (#5284 follow-ups) (#5453)
• Agent install scripts served by OneUptime itself: /install/<agent> on the reader's own host, matching its release, never GitHub (#5455)
• Resource topology as a tree of cards: no more boxes inside boxes, the way to trouble open, every product (block 50) (#5457)
• Uploaded logos inside emails, Teams in US Government clouds, a single-tenant Teams meetings app, uptime report stats by name (#5459)
• The trace view in all 17 Dashboard languages: minimap, flame graph, waterfall, span panel and signals (#5470)
• Forms and buttons, next: Change Plan on the plan picker, the criteria editor on a phone, one locked look, optional said once (#5482)
• Tables fit their card: short dates, folds by the card's width, tablet cards, compact rows (#5485)
• Status page subscribers are written to in their own language (#5501)
• Persian pages, more places: server answers and names read as written, values keep their place, popups, lists, dates and canvases follow the page (#5510)

🐛 Bug fixes

• Monitors
  • VMware datacenter alerts that never recovered, a host outage template that watched planned shutdowns, vSAN windows, and Proxmox, Ceph and Storage Array metric defaults (#5433)
  • A new infrastructure monitor keeps its default alert's criteria instead of the out-of-the-box ones (#5439)
  • An alert shows the value a summed criteria judged, not its worst minute (#5464)
  • The Kubernetes monitor form's Workload Kind follows the chosen template (#5471)
  • The Kubernetes Custom Metric tab's Workload Kind follows the chosen metric (#5474)
  • The HPA template and HPA metrics can be scoped to one HPA (#5475)
  • Unusual-count alerts on Logs and Traces keep firing while a change lasts (#5509)
  • Unusual-value alerts on Metrics monitors keep firing while a change lasts (#5513)
  • Unusual-value alerts on metrics read the right weekday, the monitor's services and a query's own scale (#5520)
  • Logs and Traces baselines read the hours the count rollups store (73576c0)
• Exceptions
  • Relative times in the reader's language, and in the sentence's own language inside a translated sentence (#5434)
  • Merged fingerprints find their survivor, and an unmerge recomputes first and last seen (#5456)
  • Merged fingerprints are matched to the exception they were merged into in the security case story, its evidence and the session replay's links (#5460)
  • An Exceptions monitor's infrastructure entity filter applies to its new and regressed exceptions (#5462)
  • What reads an exception's occurrences takes in those merged into it - the fix verifier, find_production_exceptions, the AI explanation, the spike detector, production reviews and session replays (#5465)
  • The security events on an exception's page and the Exceptions monitor form's preview take in the exceptions merged into it (#5468)
  • The exception's message keeps a reading width beside #5422's and #5416's actions (b385e48)
  • The merge dialog's description fits its limit (d6098c2)
  • An Exceptions monitor reads the clock once per evaluation (87efbbc)
• Kubernetes
  • CrashLoopBackOff Detection counts every restart, from the increase the agent sends with each sample (#5458)
  • Job Failures counts every failure, from the increase the agent sends with each sample (#5461)
  • Links to a DeploymentConfig, Route or Service keep their namespace (#5496)
  • The OpenShift verify step's SCC line leaves the translated explanation (5345e27)
• Infrastructure Monitors
  • A monitor keeps its resource by id, so a renamed cluster or a host name's case no longer empties it (#5442)
  • Request commitment leaves finished pods out, Swarm templates read tasks only, and Network Device monitors say when polls stop (#5444)
  • Ceph custom metrics per pool or daemon without scrape-counting sums, storage array thresholds in the metric's unit, and docs on health checks and timing (#5448)
• Master
  • The OpenShift pages, the Meraki read and the IoT fleet name pass compile-tests and App Test (d4420ca)
  • An OpenShift cluster's Recommendations say why it has OpenShift's alerts, and its card translates every string (7009b2c)
  • 14.0.100 cut 1's test reds - the EKS add-on's notes, the Meraki MX's firmware, private-record deletes, cached permissions and the OpenShift Events tabs (eadf1ee)
• IoT
  • Device Offline no longer fires for devices that are up, clears when a device comes back, and matches the fleet whatever its c...