Varnish stuck on most served content

[Diego Roccia]

On 03/31/2011 11:40 AM, Geoff Simmons wrote:

> On 03/31/11 11:15 AM, Hettwer, Marian wrote:
>>>>
>>>> I'm running Centos 5.5 64bit and here's my varnish startup parameters:
>>>>
>>>> DAEMON_OPTS=" -a ${VARNISH_LISTEN_ADDRESS}:${VARNISH_LISTEN_PORT} \
>>>>          -f ${VARNISH_VCL_CONF} \
>>>>          -T 0.0.0.0:6082 \
>>>>          -t 604800 \
>>>>          -u varnish -g varnish \
>>>>          -s malloc,54G \
>>>>          -p thread_pool_add_delay=2 \
>>>>                    -p thread_pools=16 \
>>>>                    -p thread_pool_min=50 \
>>>>                    -p thread_pool_max=4000 \
>>>>          -p listen_depth=4096 \
>>>>          -p lru_interval=600 \
>>>>          -hclassic,500009 \
>>>>          -p log_hashstring=off \
>>>>          -p shm_workspace=16384 \
>>>>          -p ping_interval=2 \
>>>>          -p default_grace=3600 \
>>>>          -p pipe_timeout=10 \
>>>>          -p sess_timeout=6 \
>>>>          -p send_timeout=10"
>>
>> Hu. What are all those "-p" parameters? Looks like some heavy tweaking to
>> me.
>> Perhaps some varnish gurus might shime in, but to me tuning like that
>> sounds like trouble.
>> Unless you really know what you did there.
>>
>> I wouldn't (not without the documentation at hands).
>
> Um. Many of those -p's are roughly in the ranges recommended on the Wiki
> performance page, and on Kristian Lyngstol's blog.
>
> http://www.varnish-cache.org/trac/wiki/Performance
> http://kristianlyng.wordpress.com/2009/10/19/high-end-varnish-tuning/
>
> Perhaps one of the settings is causing a problem, but it isn't wrong to
> be doing it all -- and it's quite necessary on a high-traffic site.
>
>
> Best,
> Geoff

In fact, this is one of my problems. I found ah hardly optimized 
configuration, and I'm not yet so confident to remove lines without fear :)