Brian wrote:
>
> So our multimedia cache (on a snapshot from a month or so ago) segfaulted
> out and I decided to update it. It
> so deleted those and launched squid -F
Exaplain please. What do you mean by "creamed the swap.state pretty
nicely"?
> 2001/11/30 19:40:55| aio_queue_request: Async request queue growing
> uncontrollably!
> 2001/11/30 19:40:55| aio_queue_request: Syncing pending I/O operations..
> (blocking)
> 2001/11/30 19:40:55| aio_queue_request: Synced
Fine.. not too unexpected.
> This is weird for a few reasons...
> The time is not updating while these are several minutes apart.
This is because of -F. Squid waits for -F to finish before continuing.
I have to admit that aufs haven't been tested very much with the -F flag
(well, in fact -F haven't been tested much at all). For -F operation it
is most likely better to temporary switch to using the "ufs" store
format. The two are fully compatible in the disk format, only the
internals about how I/O is performed differs.
> Squid is not posting progress reports as it rebuilds the files.
> Squid -F didn't overload the aufs queue before -- I'm not even sure it
> launched the threads before.
> Did anything change that would produce these behavior changes?
Not that I know of.
What I do know is that rebuilds without swap.state from caches created
by 2.4.STABLE1-STABLE2 is not very reliable because of one of the bugs
fixed in STABLE3 (corrupt on-disk meta data).
Regards
Henrik Nordström
Received on Sat Dec 01 2001 - 04:55:36 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:05:07 MST