test: mariadb apply retries when the certificate ask over the members' channel dies (keel#131) #132

Merged

marcos-mendez merged 1 commit from test/flaky-mariadb-cert-ask-131 into main

2026-10-09 15:20:43 -03:00

marcos-mendez commented

2026-10-09 14:30:42 -03:00

(Migrated from github.com)

The driver now retries the apply while that is the only failure (keel#131), the same treatment keel#94 gave the vip promote's etcd read. Evidence in the issue: the same signature failed three branches in one afternoon, including a tests-only branch whose product code is main's, while main itself passed before and after — a single-shot call on a lossy link, not a regression in the branches' code.

The mariadb netns job is the coverage; a green mariadb / trixie run here is the test plan.

The replica's boot apply asks the root CA's holder for the database certificate with one message. On the lossy link that ask can die and the apply fails with "did not sign the database certificate"; every dependent step is skipped, but nothing past the certificate step changed, so re-applying is safe. The driver now retries the apply while that is the only failure (keel#131), the same treatment keel#94 gave the vip promote's etcd read. Evidence in the issue: the same signature failed three branches in one afternoon, including a tests-only branch whose product code is main's, while main itself passed before and after — a single-shot call on a lossy link, not a regression in the branches' code. The mariadb netns job is the coverage; a green mariadb / trixie run here is the test plan.