fix(Propagator): Skip groupfolders root entries#58402
Conversation
Signed-off-by: provokateurin <[email protected]>
|
/backport to stable33 |
|
/backport to stable32 |
|
Might be better to have the groupfolder storage overwrite the propagator instead of having this logic in server. |
|
Right, I thought I'd have to override the whole propagateChange method, but of course it's enough to override the getParents method 👍 |
|
I remember why I this needs to be in server 😅 The propagator for the groupfolders storage wouldn't handle these, as it only propagates within its storage, so overriding its getParents won't help at all. |
|
So I could add a custom propagator to LocalRootStorage and override getParents there, but in the end the fix has to be in server, so I don't think really matters. |
|
/backport to stable32 |
|
/backport to stable33 |
|
/backport to stable31 |
This is for the "old" storage layout where all groupfolders were on a single storage.
Any change to any groupfolder caused the
__groupfoldersentry to be updated, causing locking issues on instances with lots of concurrent users.This entry is actually not used at all by the clients, because only its subfolders are mounted, therefore the etag and mtime of it don't matter.