Skip to content
Install

Monitor an app

When an app slows down or starts to fail, the Observe tab is where you find out what changed and when. It puts the app’s metrics, logs, requests, errors, and cluster events in one place, and it marks deploys and other changes on the same timeline, so you can connect a symptom to its cause.

ViewWhat it showsRequirements
MetricsTraffic, resources, pods, log volume, and deploys over time.The metrics store. Traffic panels also need a domain on the app.
LogsLive logs and past logs.Past logs need the log store.
RequestsIndividual HTTP requests.A domain on the app, the log store, and access logs.
ErrorsRepeated error log lines, grouped into one entry.Errors from replaced pods need the log store.
EventsCrashes and Kubernetes warnings.None.

The Requests view does not appear until the requirements are met.

  1. Open the app.
  2. Go to Observe > Metrics.
  3. Select a time range. The options go from Last 5 minutes to Last 7 days. To enter exact times, select Custom range.
  4. (Optional) Select By pod to split the charts by pod.

The view has these sections:

SectionPanels
Traffic and errorsRequests, Error rate, Response time
ResourcesCPU, CPU throttling, Memory, Network, Disk
PodsRestarts and out-of-memory kills, Waiting containers by reason, Pods not ready, Kubernetes warnings
LogsError and warning lines, Latest errors
DeploysDeploys, Build time, Queue wait, Where build time goes

To share the current time range and view, select the copy icon next to the time range.

A finding is a warning at the top of the Metrics view. Orchard creates a finding when one of these signals changes:

  • Error rate
  • Response time
  • Memory use, including a forecast of when the app reaches its memory limit
  • CPU throttling
  • Restarts

If a change happened in the 10 minutes before the signal changed, the finding names that change. Changes include a deploy, a change to resources or variables, and a node event.

Each finding has these actions:

  • View deploy: Opens the related deploy, if there is one.
  • Show: Zooms the charts to 15 minutes on each side of the finding.
  • Copy link: Copies a link to the finding.
  1. Open the app.
  2. Go to Observe > Logs.
  3. In Filter logs, enter text to search for.
  4. (Optional) Open the filters to select a level, a time range, or a pod.

To download logs, select the download icon. You can download the lines on screen, export every matching line, or export as NDJSON.

To follow live logs or read why a container crashed, refer to Read logs and open a shell.

  1. Open the app.
  2. Go to Observe > Requests.
  3. Filter by path, status class, method, or minimum duration.
  4. Select a request to read its details and the app logs from 30 seconds on each side.

To show only the requests from one client, select the client address in the Client column.

Orchard does not store query strings. It replaces IDs in paths with :id.

  1. Open the app.
  2. Go to Observe > Errors.
  3. Select Open, Resolved, Ignored, or All.
  4. Select a time window.
  5. Select an error to read its stack trace and the deploys it affected.

Each error has these actions:

  • Resolve or Reopen: Changes the status of the error.
  • Ignore: Hides the error from the Open list.
  • Assign to: Assigns the error to a person.

After a release settles, Orchard compares the 15 minutes before the release with the 15 minutes after it. Orchard compares response time, error rate, restarts, and memory.

Response time and error rate need at least 200 requests in each window. The result appears on the release in the Activity tab.

A result counts as worse when any of these conditions is true:

SignalCondition
Response timeThe 95th percentile rises by at least 25% and by at least 50 ms.
Error rateThe rate at least doubles and rises by at least 0.5 percentage points.
RestartsRestarts rise and reach at least 2.

A memory change of at least 30% and 64 MiB appears on the result, but it does not count as worse.

To receive a notification when a result is worse, subscribe a channel to the Got worse event (deploy.regressed).

CommandDescription
orchard observe overview <deploymentId>Print a summary for the app.
orchard observe logs <deploymentId>Read past logs.
orchard observe http <deploymentId>Read HTTP metrics.
orchard observe errors <deploymentId>Read grouped errors.
orchard observe activity <deploymentId>Read the activity feed.
orchard observe capabilities <deploymentId>List the signals available for the app.
orchard top <id>Show live pods, resources, and traffic.