Rework/improve the file scanner / file cache

[PVince81]

If we had to write it again from scratch, I have the feeling that the file scanner / cache would probably be implemented differently. The current one is mostly evolved from earlier versions.

Things that we need the file cache for:

1. storing the metadata, mostly of external storage
2. being able to detect changes on the file system (mostly for external storage)
3. be robust, support a high level of concurrency
4. propagate changes in the form of etags
5. propagate the size to compute folder sizes
6. somehow support mount points

What kind of different approach could be used to achieve these ?
Would an event system of some sorts help ?
Should the scanner be running in its own process instead of concurrently ?

@icewind1991 did you ever dream of a better file scanner ? How would that one look like ?

[PVince81]

Something I noticed is that the scanner still operates and tries to detect changes even for the home storage. It might make sense to disable the scanner completely for the home storage as the use case is that this one is ONLY used through ownCloud.

It would be good to do some performance tests and see whether it helps in any way.

Disabling the scanner would also help with concurrency issues (see #11795) where the file scanner detects FS changes in the home storage, changes done by another process which didn't have a chance yet to update the cache to match the new data.

Or... if the scanner is not disabled, allow it to run only once per user. Which means multiple scanner processes are not allowed to run at the same time for the same user.

[PVince81]

• prevent scanner concurrency

[PVince81]

Scanner concurrency issues have been solved and propagation has been reworked too.

I think we can close this.

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.