Alerts: alike problems from one place are one incident #50

Merged
jcoffey-dev merged 1 commits from feature/alert-batches into main 2026-11-16 06:03:29 +00:00
Owner

After grouping by cause, three or more alike problems from one place fold into one incident that names them, worst first, at the worst one's severity: certificates ending (by issuer, Vault mount or cluster namespace, Akamai CPS), VMs in one cluster in the same state, workloads whose pods fail the same way. Grouping's safety rules hold: a member that has paged stays on its own; a batch that has paged keeps its incident while any of it is left. Ended, weak or wrongly served certificates are never batched. Documented in formats.md, Batches.

Closes #33.

After grouping by cause, three or more alike problems from one place fold into one incident that names them, worst first, at the worst one's severity: certificates ending (by issuer, Vault mount or cluster namespace, Akamai CPS), VMs in one cluster in the same state, workloads whose pods fail the same way. Grouping's safety rules hold: a member that has paged stays on its own; a batch that has paged keeps its incident while any of it is left. Ended, weak or wrongly served certificates are never batched. Documented in formats.md, Batches. Closes #33.
jcoffey-dev added 1 commit 2026-11-16 05:52:50 +00:00
Alerts: alike problems from one place are one incident
ci / check (pull_request) Successful in 7m1s
ci / postgres (pull_request) Successful in 10m26s
8b2591982e
Twenty certificates from one Vault mount ending the same week were twenty
pages. Now, after grouping by cause, three or more alike from one place
fold into one incident that names them, worst first (eight, then a
count), at the worst one's severity:

- certificates ending, by issuer (the Certificates page's own), by Vault
  mount or cluster namespace, and Akamai CPS's;
- VMs in one cluster in the same state;
- workloads in one cluster whose pods fail the same way.

Grouping's safety rules hold: one that has paged on its own stays on its
own, and a batch that has paged keeps its incident while any of it is
left. Certificates that have ended, or are weak or served wrong too, are
never batched. Documented in formats.md, Batches.

Closes #33.
jcoffey-dev merged commit 0cdd062d47 into main 2026-11-16 06:03:28 +00:00
jcoffey-dev deleted branch feature/alert-batches 2026-11-16 06:03:29 +00:00
Sign in to join this conversation.