Monitoring apps
The metrics, alarms, logs, activity and events an app produces, and where to read each.
Metrics
The app's Metrics tab charts its own series and those of each machine. Each finished deployment
records metrics of the app (resource type app):
| Metric | Point |
|---|---|
deploy_duration_seconds |
seconds from queued to live or failed |
deploys_succeeded |
1, when a deployment goes live |
deploys_failed |
1, when a deployment fails |
The machines are compute instances; their metrics are collected under the instance and shown on
the same tab. Reading metrics needs monitoring:GetMetricData, and seeing alarms there
monitoring:DescribeAlarms.
Alarms
Any of these series can carry an alarm. An alarm on deploys_failed with statistic max, threshold
>= 1, a 30-minute period and 2 evaluation periods fires when deploys failed in two half-hours in a
row. See Monitoring.
Logs
The app's Logs tab shows what the containers print: each machine ships its lines every ten
seconds. The newest line is at the bottom and the view follows it; Pause stops following, and
Lines containing… filters. Reading logs needs monitoring:FilterLogEvents. Over the API,
GET /monitoring/api/logs?resourceId=<app id> answers newest first. A container's last lines also
appear when you select it on the Resources tab.
Deployment activity and build logs
Each deployment's Activity holds every step, written by the machine itself and by the platform, including release command output line by line; a repository build adds its Build log. See Deploy an app. Each row of Deployment history has a phase bar of how long the deployment waited, built, pushed and rolled out.
Health, sync and drift
The Resources tab shows each machine's health and sync. A machine that stays out of sync for
over a minute after a rollout is written to the app's and the machine's events (phase drift). See
Inspect an app.
Events
The default event bus receives Deployment State Change from ews.deploy when a deployment goes
live or failed, with detail.appId, deploymentId, state, error, gitCommitSha and
imageRef. For an app served through a load balancer, ews.deploy also sends
Certificate Issuance Failed when the load balancer's certificate could not be issued, and
Certificate Renewal Failed when it could not be renewed while the old one is still served. See
Event bus.
A failed, timed-out or lost scheduled-job run emails the organization's owners and admins, at most once a day per job while it keeps failing; the first failure after a success is emailed at once.