Hello,
as I was going through some upgrades during my investigation of the "transient"
queues overflows - I've noticed that my squid asserted on store.cc line 1094
which I read as a test to see if the store is in progress while the transfer is
being aborted. Sure enough I've found aborted requests for windows updates. I
have dstdomain ACL for WU sites and set quick_abort (min 4MB, max 12MB, pct
100) rules and range_offset_limit (for this ACL to 16MB). I've seen squid
downloading ot its own so I eyeballed these values on my sightsigns. I have
1Gbps download so I don't mind downloading 16MB multiple times.But it seems my
rock storage is not fast enough as I see I/O timeouts and warnings (delays I/O
requests for 72.84 seconds to obey 200/sec rate limit
etc.). Is it possible that the store is too slow and is still pending
operation while there is request abort? This rock storage is 64GB in size, slot
size 32kB, max-swap-rate 200, swap-timeout 250.
fio with randomread shows 587 IOPS (bs 32k, iodepth 1), with rwmixread 75 read
IOPS 490 write IOPS 167.
Is this related? Can I do anyzhing to avoid such asserts?
Thanks!
LL
_______________________________________________
squid-users mailing list
squid-users@lists.squid-cache.org
http://lists.squid-cache.org/listinfo/squid-users
