-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On 1/12/2014 9:44 a.m., James Harper wrote: > I've been getting squid crashes with squid-3.5.0.2-20141031-r13657. > Basically I think my cache got corrupt - started seeing > TCP_SWAPFAIL_MISS and md5 mismatches. Config is cache_dir ufs > /usr/local/squid/var/cache/squid 102400 16 256 > > It's possible that at one point I might have started 2 instances of > squid running at once... could that cause corruption? Yes, very likely. More so the longer they were both running. I see you mention segfaults below, that can also cause it for any objects in use at the time of the crash. > > And if it happens again, what sort of things should I collect to > better diagnose the problem? Firstly, Squid is expected to cope with cache corruption by cleaning up entries as it identifies them. The SWAPFAIL and MD5 details you mention above are signs that the detection at least is occuring. After any type of corruption these messages can be expected for a while with an exponential drop-off in frequency as the cache gets fixed. Only if it starts occuring with unknown cause or does not decrease in frequency is there a serious problem to attend to (usually a disk dying, Squid crashing a lot, or second Squid process started wrongly). The basic things required for bug reports (http://wiki.squid-cache.org/SquidFaq/BugReporting). That also includes investigation of the segfauls mentioned below. Plus if you can the URL, HTTP request headers in full, any access.log entry you can match up with the issue. > As I see it there are two problems: 1. that the cache got corrupt > in the first place 2. that a corrupt cache can crash squid These may in fact be the reverse with cause (2) and effect (1). When a segfault happens the details are not logged at all because the process doing the logging is the Squid which has died. Only in an assertion failure or exception error is Squid running well enough and able to log why before exiting. > > Unfortunately I did the stupid thing and deleted the cache without > taking a copy for post-mortem... the best I can do is: > > [31072.428922] squid[6317]: segfault at 58 ip 000000000061a6f9 sp > 00007fff8b9e2d40 error 4 in squid[400000+4e9000] [31654.707792] > squid[6329]: segfault at 58 ip 000000000061a6f9 sp 00007fff54358fe0 > error 4 in squid[400000+4e9000] [31783.399832] squid[6465]: > segfault at 58 ip 000000000061a6f9 sp 00007fff82af0aa0 error 4 in > squid[400000+4e9000] [31984.470507] squid[6509]: segfault at 58 ip > 000000000061a6f9 sp 00007fff028a6640 error 4 in > squid[400000+4e9000] [32178.270298] squid[6576]: segfault at 58 ip > 000000000061a6f9 sp 00007fffe64a07e0 error 4 in > squid[400000+4e9000] [32789.635935] squid[6626]: segfault at 58 ip > 000000000061a6f9 sp 00007ffff6932960 error 4 in > squid[400000+4e9000] > > addr2line -e /usr/local/squid/sbin/squid 000000000061a6f9 > /usr/local/src/squid-3.5.0.2-20141031-r13657/src/store.cc:962 > This looks like http://bugs.squid-cache.org/show_bug.cgi?id=4131. The two patched posted there seem to be getting good results. Amos -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.22 (MingW32) iQEcBAEBAgAGBQJUfEVFAAoJELJo5wb/XPRjgVkH/1gj4utqZanu9jtLFEaBuV9Q c4K+pw7grL8s1CH8vKvyNZnhSiDi7FsPPaG1RQTkJplhsqswZ2rUcLkVwAEHb2Ug cS4uH9y5nN/M+O7yqmx/29JS1ITaXnR2ooy8PctKZoYqizEIz6UhDDd2vFuKiPFJ rlhP+gvd8fDACtZgWLnojl6OmrFXmD0RyxZE0r8Y5wQyzIkbqveJfHzRcl7hkZJh xJLPfiakK0RBHQSEDRJg/Jui8hv2UeaqGd/YJcF+XJZW6USY6tB8sVnwd8zir6Aw q2VVbofu2YRn7RJmUrwwppvbmQ+j9ykRS5VMFkJrDVGrf0RohIqn1d7OO3pNJu0= =DOp+ -----END PGP SIGNATURE----- _______________________________________________ squid-users mailing list squid-users@xxxxxxxxxxxxxxxxxxxxx http://lists.squid-cache.org/listinfo/squid-users
