ydb-platform / nbs

Network Block Store
Apache License 2.0
50 stars 14 forks source link

issue-1350: simplified node creation in follower code, added a test for errors received from follower in this case, fixed DupCache entry commitment in CreateHandle, made TabletProxy immortal, deleted inactive pipe tracking and closure in TabletProxy, added some comments, fixed logging a bit #1511

Closed qkrorlqr closed 4 days ago

qkrorlqr commented 6 days ago

DupCache entry commit was simply forgotten for the multi-tablet mode in my previous PRs. And as for TabletProxy - we send various requests to IndexTablets via TabletProxy and there is some weird logic in IndexTablet which sends EvPoisonPill in response to some previously sent CreateSessionRequests - this logic controls the lifecycle of the CreateSessionActor and is quite fragile so I'm not planning to change it right now. But this logic might kill TabletProxy in some cases which renders the filestore-server instance almost unusable. VolumeProxy in NBS is already immortal - making TabletProxy immortal in a similar fashion. As for the inactive pipe tracking logic - it's not really needed and it messes with the leader->follower session creation logic (followers destroy sessions after pipe destruction which breaks some scenarios). Deleting that unneeded logic is the most straightforward solution.

1350

github-actions[bot] commented 6 days ago

[!NOTE] This is an automated comment that will be appended during run.

:green_circle: linux-x86_64-relwithdebinfo: all tests PASSED for commit 4bd34446ed9ce8c71d9f1167d4a87fc7ba9e850d. TESTS PASSED ERRORS FAILED SKIPPED MUTED?
1639 1639 0 0 0 0
github-actions[bot] commented 5 days ago

[!NOTE] This is an automated comment that will be appended during run.

:red_circle: linux-x86_64-relwithdebinfo: some tests FAILED for commit 1c127ac254cb7bfb86d3ed0419a02f442c522772. TESTS PASSED ERRORS FAILED SKIPPED MUTED?
1641 1639 0 2 0 0
qkrorlqr commented 4 days ago

TEST NAME ELAPSED STATUS LOG cloud/filestore/tests/fs_posix_compliance/qemu-kikimr-nemesis-test/[0/7] chunk chunk 0.000s FAIL DIR | LOG | STDERR | RECIPE_STDERR | RECIPE_STDOUT cloud/filestore/tests/fs_posix_compliance/qemu-kikimr-nemesis-test/test.py.test_posix_compliance[flock] 9m 24.309s FAIL(TIMEOUT) DIR | LOG

флап, чиню тут https://github.com/ydb-platform/nbs/pull/1520