On Mon, 13 Dec 2010 17:05:39 -0800, Michael Leong <Michael.Leong@xxxxxxxxxxxxxxxx> wrote: > Hi, > > I'm using squid 2.x because the site says that's the recommended version > for high throughput caches. The disk calculation bug seems to only > occur on NFS/Netapp squid stores. We're using the identical builds w/ > regular harddisk stores and this problem doesn't exist. > > - do you recommend upgrading 3.x for caches demanding high throughput? It is up to you whether the drop is acceptable. All we can say is that you can expect 3.1 to be around 15% slower than 2.7 on a request/second throughput. > - can I upgrade to 3.x and keep my existing cache? COSS is not reliably working (yet) in 3.x. The other cache types are upgraded automatically by squid. Amos > > thx > > On 12/13/2010 5:01 PM, Amos Jeffries wrote: Re: severe > disk usage calculation bug > > On Mon, 13 Dec 2010 14:10:21 -0800, Michael Leong wrote: > > Hi, > > we're currently running the squid store off a NetApp NFS. According to > > df, it says our cache store is using 140GB. However, when I run the > > disk usages query via > > > > snmpwalk -c public -v2c localhost:3401 .1.3.6.1.4.1.3495.1.1.2.0 > > > > Squid thinks it's using about 1TB of storage. This is causing serious > > issues because squid started to purge my cache. Any ideas on why and > > how to solve this? > > Sounds like http://bugs.squid-cache.org/show_bug.cgi?id=3068 [1]. > Fixed in 3.1.10 which will be out shortly, or the 3.1.9 daily bug fix > bundles. > > Amos >
