More on this, I now get this quite often: 2005/12/20 17:00:56| ctx: exit levels from 1 down to 0 2005/12/20 17:00:56| assertion failed: mem_node.cc:65: "n->write_pending" After which Squid restarts. The version I'm trying out now is PRE3-20051219. Should I open a bug of this? Tuukka > -----Original Message----- > From: Laurikainen, Tuukka [mailto:t.laurikainen@xxxxxxxxxxxxxx] > Sent: Tuesday, December 20, 2005 1:28 PM > To: Squid Users > Subject: parse reply headers problem > > Hi, > > Squid-3.0PRE3 as a reverse Proxy. After changing the cache from ufs to > aufs and increasing the cache_mem size, one of the pages published > produces these errors on cache.log: > > could not parse headers from on disk structure! > clientReplyContext::sendMoreData: Unable to parse reply headers within a > single HTTP_REQBUF_SZ length buffer > > From the store.log I can only see memory events related to this page: > > RELEASE -1 FFFFFFFF > > I fixed this by regenerating the cache (delete cache, generate with > squid -z). Would there have been some other way to recover from the > situation? > I'm not sure what really caused all this, as the ufs -> aufs change > should not affect the cache data, and the increase in cache_mem was not > exaggerated neither (256MB on 1GB system). However I returned to a lower > value just in case. > > Regards, > > Tuukka
