Avoid locking on restore/delete and limit concurrent requests#36465
Avoid locking on restore/delete and limit concurrent requests#36465juliusknorr merged 3 commits intomasterfrom
Conversation
|
Edit: that doesn't work as they don't have a common parent but are instead named "$filename.vXXX.dXXX" |
|
Note that I'm not sure if the scan logic is still needed, I know that in the past we didn't actively maintain the cache outside of |
I had the same assumption in https://github.com/nextcloud/server/pull/28438/files#r691270467 so should we just drop it? |
|
Deleting and restoring versions seems to still work fine with that code part removed. |
4f9d903 to
9ebe372
Compare
|
Do we want to increase the concurrency a bit if we remove the scan? |
…ocking in the backend Signed-off-by: Julius Härtl <jus@bitgrid.net>
…rsions from trash Signed-off-by: Julius Härtl <jus@bitgrid.net>
9ebe372 to
7f913de
Compare
|
Increased to 4 which seems a reasonable value for now. |
|
/compile |
Signed-off-by: nextcloud-command <nextcloud-command@users.noreply.github.com>
|
/backport 7f913de to stable25 |
|
/backport 7f913de to stable24 |
Summary
This is a frontend safeguard to avoid running into backend file locks when deleting or restoring multiple files. Even after https://github.com/nextcloud/server/pull/28438/files#diff-5450a6734b54a8b5e2c2f7521098ba6d05993e8bbebffeee140e93ec349e4127R978 this could still happen if the retry timeout was exceeding 15 seconds.
Checklist