osd: Test peering flush batches waiting for the PG history #132

Merged
vitalif merged 1 commits from koolkhel/vitastor:peering-flush-minsize-tests into master 2026-11-19 15:30:16 +03:00
Contributor

Description

Two tests for paths of #131 that no test reached:

  • test_peering_flush_waiting_batch_repeer: EC 2+1, a peering flush batch waits for the history with the holders to be reported, and OSD 3 drops out before the report lands. The batch is dropped without sending any stable or rollback op, and the PG peers again. Without the continue_pg() in the drop branch of on_change_pg_history_hook() the PG stays repeering and the test fails.
  • test_peering_flush_replicated_history: replicated pool, v2 stable on OSD 1 and unstable on OSD 2. OSDs 1 and 2 are in the history before the flush is sent. It fails if replicated pools are skipped when adding the holders.

Tested: osd_test, all pass on current master.

Contributor Name

Yury Luneff (MIND Software LLC, https://mindsw.io)

CLA

### Description Two tests for paths of https://git.yourcmc.ru/vitalif/vitastor/pulls/131 that no test reached: - `test_peering_flush_waiting_batch_repeer`: EC 2+1, a peering flush batch waits for the history with the holders to be reported, and OSD 3 drops out before the report lands. The batch is dropped without sending any stable or rollback op, and the PG peers again. Without the `continue_pg()` in the drop branch of `on_change_pg_history_hook()` the PG stays repeering and the test fails. - `test_peering_flush_replicated_history`: replicated pool, v2 stable on OSD 1 and unstable on OSD 2. OSDs 1 and 2 are in the history before the flush is sent. It fails if replicated pools are skipped when adding the holders. Tested: `osd_test`, all pass on current master. ### Contributor Name Yury Luneff (MIND Software LLC, https://mindsw.io) ### CLA - [x] I accept Vitastor CLA agreement: https://git.yourcmc.ru/vitalif/vitastor/src/branch/master/CLA-en.md
koolkhel added 1 commit 2026-11-19 14:17:13 +03:00
A peering flush that stabilizes the target version of an object waits
until the OSDs holding that version are reported in the PG history.

Test that such a waiting batch is dropped without sending anything when
the PG repeers before the report lands, and that the PG then peers
again. Also test that the holders go into the history for a replicated
pool, before the flush is sent.
vitalif merged commit 7c76601c45 into master 2026-11-19 15:30:16 +03:00
Sign in to join this conversation.
No Reviewers
No Label
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: vitalif/vitastor#132
No description provided.