On 17/05/2013 8:37 p.m., lantuin wrote:
Hi, we have an instance of squid - running version 3.1.8 - that suddenly has gone in FATAL error. The reason that I found in cache.log is this: 2013/05/13 15:42:46| mem_hdr::write: writeBuffer: [341,553) 2013/05/13 15:42:46| mem_hdr::debugDump: lowest offset: 0 highest offset + 1: 342. 2013/05/13 15:42:46| mem_hdr::debugDump: Current available data is: [0,342) - . 2013/05/13 15:42:46| storeDirWriteCleanLogs: Starting... 2013/05/13 15:42:46| WARNING: Closing open FD 11 2013/05/13 15:42:46| Finished. Wrote 41977 entries. 2013/05/13 15:42:46| Took 0.04 seconds (1085490.42 entries/sec). FATAL: Attempt to overwrite already in-memory data. Preceeding this there should be a mem_hdr::write output that lists the attempted write, and the currently present data. Please get a 'backtrace full' from this error - using the generated core, and file a bug report with the squid developers including the last 10 lines of cache.log and the backtrace. Note: we have 6 instances of squid handling heavy load under a load balancer. The strange thing is that ALL 6 istances in the same identical moment presented this FATAL ERROR. I've found in internet another person with the same error but I think he didn't resolve the issue (http://www.squid-cache.org/mail-archive/squid-users/201108/0024.html). Somebody please can help me?
Strange that it should be synchronised like that. Like the message says, please find the backtrace. details on that can be foudn in the wiki at http://wiki.squid-cache.org/SquidFaq/BugReporting if you need a how-to. Since this version of Squid is quite old now please also carefully check for *closed* bugs before reporting.
You would also be well served by upgrading to the current releases on general principles. As of now the current squid release is squid-3.3.4 and resolves a few security vulnerabilities, and a great many bugs in 3.1.
Amos
