On 2/6/13 2:22 PM, Jan-Frode Myklebust wrote:
> I have a reverse proxy that's serving about 1.5 million requests/day,
> and about 20 GB data (and I have a 30 GB cache volume for ATS). Looking
> trough my squid log, I see several HITs that are taking several seconds
> to complete:
>
> 1360166550.001 28898 188.x.y.z TCP_HIT/200 573359 GET http://content.example.com/nuxeo/site/example/tv/partner/1000
- NONE/- text/xml -
> 1360134095.057 29304 79.x.y.z TCP_HIT/200 1550 GET http://content.example.com/nuxeo/nxfile/default/d2128a17-e6c3-47ca-bc41-4865c07f662c/tvch:xsiteLogo/NRK2_DB.png
- NONE/- image/png -
>
> What can explain ~30 elapsed time to complete HITs of 500KB and 1.5 KB ?
> Is it affected by slow clients?
>
Typically slow clients in my experience. I'd suggest you enable the "slow
log" feature, put it at a threshold of e.g. 25s, and it will log some
information about where time is being spent. Do you have any timeout set at
around 30s ?
-- Leif
|