BugZero found this defect 41 days ago.
Data sources
All data on this page is proprietary to BugZero® or gathered from public sources
4/8/2024
MongoDB Server
No affected releases provided.
No fixed releases provided.
The test attempts to restart nodes during a resharding operation. This can result in the following situation: resharding operation chooses Timestamp T=0 replica set advances stable timestamp to T=1 secondary restarts and sets the minimum valid snapshot timestamp to be the stable timestamp resharding operation attempts to scan the config.transactions table at T=0 on the secondary resharding receives a SnapshotUnavailable error since the collection instances do not match because we created a new PIT collection instance.
JIRAUSER1264163 commented on Mon, 8 Apr 2024 13:29:56 +0000: I suppose we could do that here since it's definitely a test-only issue. I have to say though that making failpoints acting more out of line from what the server actually does normally is a bit odd to say the least. Something to note also is that the minVisible parameter hasn't existed since SERVER-68270 so today this failpoint isn't doing anything. max.hirschhorn@10gen.com commented on Mon, 8 Apr 2024 12:57:57 +0000: jordi.olivares-provencio@mongodb.com, could we instead make the PIT CollectionCatalog aware of the setMinVisibleForAllCollectionsToOldestOnStartup failpoint for how it sets minValid in addition to minVisible? This SnapshotUnavailable issue in testing was meant to be worked around by the changes from SERVER-59923.