You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Apr 21, 2023. It is now read-only.
1. Configure Apache with mod_pagespeed and multiple vhosts
2. Each vhost should have it's own FileCachePath or CacheFlushFilename
3. Don't have UsePerVHostStatistics turned on
If you touch the cache flush file for one cache, only that cache should be
flushed. Instead all caches are flushed.
This happens because in SystemServerContext there's a statistic
kCacheFlushTimestampMs which is, with UsePerVHostStatistics disabled, shared by
all vhosts.
It's actually worse that this, in that the cache is only over-flushed
sometimes. FlushCacheIfNecessary either checks the mtime of the cache flush
file or the contents of the kCacheFlushTimestampMs (depending on elapsed time
since the last check of the filesystem) but not both. This means that if you
only get in a request every 10s and have a 5s CacheFlushPollIntervalSec then
you'll never see this problem because you'll only ever use the on-disk mtime.
If you start getting requests every 1s, however, you'll use the statistic
4/5ths of the time and if another virtual host is being flushed often then
you'll be flushed with it.
Why do we need a statistic for this? Why can't each process just check the
timestamp itself?
Original issue reported on code.google.com by jefftk@google.com on 1 Apr 2013 at 10:14
The text was updated successfully, but these errors were encountered:
Original issue reported on code.google.com by
jefftk@google.com
on 1 Apr 2013 at 10:14The text was updated successfully, but these errors were encountered: