Thursday 08 October 2009 3:02:11 am
I found out what is the problem! At least in my case where we DON'T have the noatime attribute set for the partition. There is an ini setting in site.ini under ContentSettings:
# maximum cache generation time. If a file stays in generation mode for more than
# this value in seconds, it is considered timed out (process died, for instance),
# and generation is taken over by the requesting process
CacheGenerationTimeout=60
So with stale cache, if the cache is not generated in 60 seconds, the cache generation process starts over again. It seems that for heavy sites 60 seconds is not enough, so the regeneration of cache repeats and repeats all over again. Now I have set this limit to 120 sec for my site and it's working OK, and there are no timeout messages in the log.
But be careful not to set this setting to high, because it could generate a lot of web server's processes just waiting for cache generation. You should set this value to the maximum time a cache needs to be generated. One way to detect the right value is by increasing the value by 10 sec, testing the site and looking at the log for cache generation timeout messages. And then repeating this until you find the value which doesn't generate timeout messages. P.S. The REAL problem is that CacheGenerationTimeout setting is not mentioned anywhere on ez.no. It's only described in site.ini. This setting is very important and should be mentioned in the docs. Please do it.
--
Nothing is impossible. Not if you can imagine it!
Hubert Farnsworth
|