On fre, 2007-11-23 at 09:55 +0900, Adrian Chadd wrote: > I've been fiddling with Squid-2.HEAD and google maps, and I've noticed that > weird stuff happens. > > Here's the log entries: > > 1195739570.596 SWAPOUT 00 00034F1B BEAF584D4381EE1DCAF38A162BAA0224 200 > 1195739570 1103259488 1227275570 image/jpeg 11389/11389 GET > http://kh2.google.com.au/kh?n=404&v=23&t=ttrtq > 1195777355.041 RELEASE 00 00034F1B 748D838AAD5499315AEE444C306BE06C 200 > 1195739570 1103259488 1227275570 image/jpeg 11389/-243 GET > http://kh2.google.com.au/kh?n=404&v=23&t=ttrtq > 1195777356.133 RELEASE -1 FFFFFFFF C208FB77DE2B773EFFBCF7DA928D1DF7 200 > 1195777354 1103259488 1227313354 image/jpeg 11389/11389 GET > http://kh2.google.com.au/kh?n=404&v=23&t=ttrtq > > Now, the log entries in access.log so TCP_MISS, then TCP_REFRESH_MISS. > > I'll sort out the revalidation issue, but I note that there's a negative > object > size there - that -243 is in store_log.c: > > mem->inmem_hi - mem->reply->hdr_sz, > > Any insights into that?
I would guess the object was purged from memory before RELEASE, and that our changes in the store client wrt how headers on swapins is handled makes a memobject to be attached to the StoreEntry again but without any data.. Regards Henrik
signature.asc
Description: This is a digitally signed message part