On 27/09/2016 3:22 a.m., Jasper Van Der Westhuizen wrote: > > Is there anything that you guys can suggest I do around the cache? > Should I try a different store type? A different filesystem type > perhaps? > > > > If your store has a configuration knob that effectively limits disk > writing rate, then use it to limit that rate to avoid overflowing the > queue. > > You can also consider using rock store that has an explicit write > rate limiting option (but comes with other problems that may or may > not affect your setup). > > Adding more physical disk spindles helps, of course. > Just to be clear since you mentioned working with virtual services. By "spindle" Alex refers to individual physical HDD devices underneath the whole VM storage setup. 'Disk' a the VM level, even at the RAID level might refer to multiple or overlapping (shared) 'spindles'. Unless your AUFS/UFS/diskd cache_dir are limited to being on one "spindle" HDD, they can cause a shared HDD controller to fill up / overload and adding more cache_dir just makes that particular problem worse. Amos _______________________________________________ squid-users mailing list squid-users@xxxxxxxxxxxxxxxxxxxxx http://lists.squid-cache.org/listinfo/squid-users
