Skip to content

Add steps for submitting Sourcegraph metric dumps to support

Warren Gifford requested to merge sg/metrics-dumps into master

Created by: slimsag

The goal of this PR is to add steps we can link customers with performance/instability issues to such that we can then inspect the Prometheus metrics interactively ourselves to identify the root cause of their issue. The process looks like this:

  1. Send customer a link to this docs section, and a share a private Google Drive folder with them so they can upload (the sometimes quite large) metrics dump.
  2. Once it is uploaded, download it and copy the extracted metrics dump into ~/.sourcegraph/prometheus.
  3. Run the same Sourcegraph version as them, investigate via the local Grafana instance / by running Prometheus queries directly.

In the future we can:

  1. Provide the same steps for Kubernetes
  2. Modify dev/run-server-image.sh to accept a PROMETHEUS_DUMP=somedump.tgz env var so the process is seamless for us.

Merge request reports