-
Type:
Bug
-
Resolution: Done
-
Priority:
Major - P3
-
None
-
Affects Version/s: 8.0.4
-
Component/s: None
-
ALL
-
-
(copied to CRM)
-
None
-
None
-
None
-
None
-
None
-
None
-
None
On Flex clusters specifically, change streams can report an `ns.db` value of for example `67b83e86cd20730f1e766dde_ps` instead of `ps`. This generally only happens when changes were made while the client was offline, and then the client catches up using `resumeToken` or `startAtOperationTime` - then the events made while offline has the incorrect `ns.db` value, while the new ones have the correct value.
This only happens when watching on a single database. When watching on multiple databases, I could not reproduce the same issue. So it does have a simple workaround at least since I know what the database should be, but it does still appear to be a bug.
This also appears to be specific to Flex clusters. I have not been able to reproduce on dedicated or self-hosted clusters. I haven't tested this with free clusters yet.
In my specific example here I'm watching for events on a single collection, so the ns value is not important. However, I used this in production for a use cases that could potentially get events on multiple different databases and collections, where the exact value is important.