Thanks Ken, I updated the vm.min_free_kbytes to 128M and increased the cli_timeout to 40.
No there was no errors from any parts of the HW (ram,disk ...) Jef Le 22/04/11 03:06, « Ken Brownfield » a écrit : > It was likely killed because the child process didn't respond to the parent > process within cli_timeout seconds. > > Two suggestions: > * increase vm.min_free_kbytes to 64M or 128M (I use the latter) > * increase cli_timeout to 40-60 seconds > > Then record 'vmstat' and 'ps' output, so you can go back and see what was > happening on the machine at the time the child is killed. For example, did > something start swapping 1G of RAM, and swap thrash caused the child to be > slow to respond)? Was there an ECC error? Etc. > > My guess is that the two settings above will put you in a better place. Also, > good to make sure your lru isn't too high, but it looks like you're using the > (fine) default. > -- > kb > > > > On Thu, Apr 21, 2011 at 15:59, Jean-Francois Laurens > <[email protected]> wrote: >> Please do not overlook the object of the email which is actually ³Varnish >> child killed². >> >> For the explanation of the hitratio: some tests are running and half of the >> misses are requests with cachekillers on purpose. >> >> Would anybody please help on understanding why this child process get killed >> ? >> >> >> Le 21/04/11 19:41, « Simon Lyall » a écrit : >> >>> On Thu, 21 Apr 2011, Jean-Francois Laurens wrote: >>>> > 2. hitratio >>>> > My hit ratio is pretty low around 50%. >>>> > It used to be around 70% with the version 2.0.6 and dropped down to 50%. >>>> > I can¹t explain it as what was changed was only : >>>> > In vcl_fetch, replace obj by beresp >>>> > Replace calls to pass or deliver by return(deliver) or return(pass) >>> >>> Is this for http://www.tsr.ch/ ? >>> >>> If your hit ratio is 60% with a million objects then something is very >>> wrong. You have enough images there you should be well over 90% hits. >>> >>> Run "varnishlog -b" >>> >>> and see what your cache misses are ( or do look in your backend logs ). >>> >>> >>> Jean-Francois Laurens >>> Ingénieur Système Unix >>> Resources et Développement >>> Secteur Backend >>> RTS - Radio Télévision Suisse >>> Quai Ernest-Ansermet 20 >>> Case postale 234 >>> CH - 1211 Genève 8 >>> T +41 (0)58 236 81 63 <tel:%2B41%20%280%2958%20236%2081%2063> >>> >>> >>> _______________________________________________ >>> varnish-misc mailing list >>> [email protected] >>> http://www.varnish-cache.org/lists/mailman/listinfo/varnish-misc >>> >>> >>> >>> _______________________________________________ >>> varnish-misc mailing list >>> [email protected] >>> http://www.varnish-cache.org/lists/mailman/listinfo/varnish-misc >>> >>> Jean-Francois Laurens >>> Ingénieur Système Unix >>> Resources et Développement >>> Secteur Backend >>> RTS - Radio Télévision Suisse >>> Quai Ernest-Ansermet 20 >>> Case postale 234 >>> CH - 1211 Genève 8 >>> T +41 (0)58 236 81 63 >>>
_______________________________________________ varnish-misc mailing list [email protected] http://www.varnish-cache.org/lists/mailman/listinfo/varnish-misc
