Fix flaky keepstore test
Keepstore test failed on a branch changing an unrelated thing:
#4 Updated by Tom Clegg about 2 months ago
- Status changed from New to In Progress
This was already fixed incidentally in #17698 by changing (*RRVolumeManager)AllReadable() to return volumes sorted by priority and then UUID, instead of indeterminate order as before. The index handler processes the volumes in the same order AllReadable() returns them, so the test is now reliable.
Test results today:
32x at 87b0f117ff1ce917d7301f8a6304e931a189e48d (version jenkins tested in the given example) → 6 failures
32x at 2c2153daa103c35fe12468a853636f8125785a7e (head of main branch today) → 0 failures
#7 Updated by Tom Clegg about 2 months ago
- Status changed from Resolved to In Progress
Another one still happening on main:
FAIL: mounts_test.go:21: HandlerSuite.TestMounts
#8 Updated by Tom Clegg about 2 months ago
Doh. Fixed by sorting vm.mounts the same way as vm.readables and vm.writables. This time I checked that all keepstore tests pass 32x in a row.