Purge of FileCache in 6.5.1 nightlies after 10/6 takes excessive time for large collections;
Long enough to stall slimserver, and starve SB clients. :(


[QUOTE=andyg;147724]Any version of 6.5.1 after 10/6.[/QUOTE]

On my system(s), Windows XP "enough" space on C:\ but with a large collection (300GB \ 500GB +) of tagged flacs with album art Typically 300x300 - ish, on 6.5.1 nightlies after this fix, I've noticed the following.

At hourly intervals (consistent with the cache purge fix), Lots of disk activity, some Slim CPU activity, (indicating its I/O bound) lasting for between say 3-8 minutes. This stalls all clients of slimserver (indicating slimserver is busy on its main thread?).

Observable effects.

Playing Slim devices:
Buffer empties, not refreshed, music stops. Unresponsive to any command until disk activity finished. So thats only about 90% uptime on an hourly basis when listening :(

Non playing slim devices running clock screensaver:
Clock does not update - giving good feedback of when issue started - and how long it lasts.

Anybody else experiencing this issue?

Is a "one size fits all" solution appropriate for all cases?
Would some logging and recording help?
(e.g. starting purge... purge ended ... took X mins)
Log the elapsed time for a purge, and if its longer than some threshold - say the buffer size on a SB, defer the next purge, or  take into account free space on the disk, or whatever).

Moving cache purge to separate thread? ( is this possible?, what about cache coherency etc)

Any more info you need?

I've opened this as Bug