I think it's due to dns. Here was the squid manager output:
Median Service Times (seconds) 5 min 60 min:
HTTP Requests (All): 8.68295 2.37608
Cache Misses: 10.20961 0.03066
Cache Hits: 8.22659 2.79397
Near Hits: 0.00000 0.00000
Not-Modified Replies: 0.00000 0.00000
DNS Lookups: 10.60242 9.70242
ICP Queries: 0.00000 0.00000
Does squid still use dns for reverse proxy requests? All my requests
goes to http://cache-int/, but cache-int is not on /etc/hosts nor on
DNS. I have 1 orginal-server defined and is used as the default, so
shouldn't squid just goto the backend w/o dns lookups?
thx,
mike
At 03:10 PM 6/6/2008, Henrik Nordstrom wrote:
On fre, 2008-06-06 at 14:38 -0700, leongmzlist wrote:
> My cache performance is acting strange; I'm getting extremely high
> tcp_hit times for cached objects:
>
> 1212787643.465 50343 10.2.7.22 TCP_HIT/200 19290 GET http://cache-int/....
> 1212787737.740 15212 10.2.7.25 TCP_HIT/200 11511 GET http://cache-int/....
>
>
> Those high times comes in bursts. Eg: bunch of high response time
> will come followed by a normal response times. Normal response times
> are sub 100ms
Could be cache validations. Some times TCP_HIT is logged when it really
should have been TCP_REFRESH_HIT. This can happen if the object uses
Vary if I remember correctly.
Another possibility is if the Squid serer is swapping, causing Squid to
delay everything waiting for swap activity.
A third possibility is if you have ACLs which may cause delays, such as
DNS dependencies or external acl lookups.
Regards
Henrik
