Skip to content

docs(admin): document classic and native histograms - #1239

Draft
pbleser-oc wants to merge 2 commits into
opencloud-eu:mainfrom
pbleser-oc:feat/1235-prometheus-histograms
Draft

pbleser-oc wants to merge 2 commits into
opencloud-eu:mainfrom
pbleser-oc:feat/1235-prometheus-histograms

Conversation

@pbleser-oc

Copy link
Copy Markdown
Member

Fixes #1235 (part of opencloud-eu/opencloud#3817)

Note

This builds on #1238, which adds the page "Monitoring with Prometheus"; its commit is included here until it is merged. Only the last commit, bd07a58, belongs to this PR: see the diff of that commit.

It documents OpenCloud PRs that are not merged yet: opencloud-eu/opencloud#3825 (classic and native histograms), opencloud-eu/opencloud#3827 (deprecated Summaries) and opencloud-eu/opencloud#3828 (bucket sets). It stays a draft until they are merged.

Changes

Updates the page "Monitoring with Prometheus" for the changes to the histograms:

  • Both forms: the duration histograms are exposed as classic and native histograms at the same time, with at most 100 native buckets.
  • Switching to native histograms without losing data with always_scrape_classic_histograms, in four steps, including a warning for setups that already have native histograms enabled when they upgrade OpenCloud.
  • PromQL examples for both forms side by side.
  • Bucket sets: the four shared sets, and a caution box with the histograms whose bucket boundaries change.
  • Deprecated Summaries of invitations and webfinger, and the histograms to use instead (seconds instead of microseconds).
  • The list of duration histograms with their bucket sets and native form.

Checks

pnpm build --locale en, prettier --check and markdownlint-cli2 pass for the changed files (the broken anchors that the build reports are on existing developer pages).

Add the admin page "Monitoring with Prometheus", as OpenCloud works
today: where OpenCloud exposes its metrics (debug servers, one endpoint
per process, the OpenCloud Compose monitoring files, the debug token),
a Prometheus scrape configuration, and the duration histograms: classic
histograms with their buckets and PromQL examples, the native-only
request duration histogram of the proxy with its classic fallbacks, how
to enable native histograms per Prometheus version, and a list of all
duration histograms.

Fixes opencloud-eu#1237
Update the page "Monitoring with Prometheus" for the changes to the
histograms in opencloud-eu/opencloud#3817:

- the duration histograms are exposed in both classic and native form
  (opencloud-eu/opencloud#3825), with at most 100 native buckets;
- how to switch to native histograms without losing data, with
  always_scrape_classic_histograms, including a warning for setups that
  already have native histograms enabled when they upgrade;
- PromQL examples for both forms side by side;
- the shared bucket sets and the histograms whose bucket boundaries
  change (opencloud-eu/opencloud#3828);
- the deprecated latency Summaries of invitations and webfinger, and the
  histograms to use instead (opencloud-eu/opencloud#3827);
- the list of duration histograms with their bucket sets and native form.

Fixes opencloud-eu#1235
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

Status: Qualification

Development

Successfully merging this pull request may close these issues.

Admin docs: Prometheus native and classic histograms

1 participant