Strange caching behaviour with certain objects

From: Ernest Yik <[email protected]>
Date: Fri, 6 Jun 1997 22:37:29 +0800 (HKT)

Hi Squid managers,

Can anybody explain the following behaviour?

Case 1 :

Jun 6 21:27:01 4208 client TCP_MISS/200 1071 GET http://bart.prod.oclc.org:3050/images/fssrchg.gif - FIRST_PARENT_MISS/squid1 image/gif
Jun 6 21:27:12 4641 client TCP_MISS/200 1071 GET http://bart.prod.oclc.org:3050/images/fssrchg.gif - PARENT_UDP_HIT_OBJ/squid1 image/gif
Jun 6 21:27:34 3703 client TCP_HIT/200 1071 GET http://bart.prod.oclc.org:3050/images/fssrchg.gif - NONE/- image/gif
Jun 6 21:27:58 5434 client TCP_HIT/200 1071 GET http://bart.prod.oclc.org:3050/images/fssrchg.gif - NONE/- image/gif

client requests an object which is not in cache, got it from a parent
squid, but doesn't cache it. It caches the second request of the same
object, which is served from the cache of the parent. Subsequent
requests got TCP_HIT. Why doesn't the object got cached the first time
but the second time?

Case 2 :

Jun 6 21:30:30 24215 client TCP_MISS/200 892 GET http://medusa.prod.oclc.org:3050/images/fs_help.gif - FIRST_PARENT_MISS/nsproxy image/gif
Jun 6 21:30:42 4334 client TCP_MISS/200 890 GET http://medusa.prod.oclc.org:3050/images/fs_help.gif - FIRST_PARENT_MISS/squid2 image/gif
Jun 6 21:30:53 5167 client TCP_MISS/200 890 GET http://medusa.prod.oclc.org:3050/images/fs_help.gif - PARENT_UDP_HIT_OBJ/squid2 image/gif
Jun 6 21:31:03 3579 client TCP_HIT/200 890 GET http://medusa.prod.oclc.org:3050/images/fs_help.gif - NONE/- image/gif
Jun 6 21:31:20 3284 client TCP_HIT/200 890 GET http://medusa.prod.oclc.org:3050/images/fs_help.gif - NONE/- image/gif

client requests an object which is not in cache, got it from a parent
netscape proxy, but doesn't cache it. Second request fulfilled by
another parent squid, but still doesn't cache it. It caches the third
request of the same object, which is served from the cache of the
parent. Subsequent requests got TCP_HIT. Seems that this squid only
caches the object if it is returned from a PARENT_UDP_HIT_OBJ but not a
FIRST_PARENT_MISS?

Case 3 :

Jun 6 22:12:04 23621 client TCP_MISS/200 2499 GET http://bart.prod.oclc.org:3050/images/oclctop.gif - FIRST_PARENT_MISS/squid2 image/gif
Jun 6 22:12:05 124 client TCP_MISS/200 2499 GET http://bart.prod.oclc.org:3050/images/oclctop.gif - FIRST_PARENT_MISS/squid2 image/gif
Jun 6 22:12:06 143 client TCP_MISS/200 2499 GET http://bart.prod.oclc.org:3050/images/oclctop.gif - PARENT_UDP_HIT_OBJ/squid2 image/gif
Jun 6 22:12:07 39 client TCP_HIT/200 2499 GET http://bart.prod.oclc.org:3050/images/oclctop.gif - NONE/- image/gif
Jun 6 22:12:08 64 client TCP_HIT/200 2499 GET http://bart.prod.oclc.org:3050/images/oclctop.gif - NONE/- image/gif

Actually the same case, but the first 3 requests show that there maybe a
similar problem with the parent squid (I'm not alone?).

Tested with both 1.1.10 and 1.NOVM.10 under Linux 2.0.30, same
behaviour. Parent squids (squid1 and squid2) are both 1.1.10. I can
only reproduce this problem with the http://*.prod.oclc.org:3050/* URLs.
Most (if not all) other URLs behaved normally. So this may be a
HTTP-server side problem. Or maybe it's my configuration? Any hints?

I can set a higher debug_option and send you a copy of the log if it can
help. Thanks.

Regards,
Ernest Yik

-- 
 * Origin: VAXlab : Home of Ernest * Hong Kong * Email: ernestyik@cuhk.edu.hk
Received on Fri Jun 06 1997 - 07:40:00 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:35:28 MST