If you're running an FC3 mail server with Postfix that's quite busy, please check your mail logs. Some Postfix daemons may crash and then get restarted, but the whole thing is not evident unless you dig through the logs: grep panic /var/log/maillog* You may see errors such as: Apr 18 12:19:14 XXXXgate postfix/smtp[30810]: panic: myfree: corrupt or unallocated memory block Apr 18 12:19:15 XXXXgate postfix/qmgr[29035]: warning: transport smtp failure -- see a previous warning/fatal/panic logfile record for the problem description Apr 18 12:21:17 XXXXgate postfix/qmgr[29035]: warning: transport smtp failure -- see a previous warning/fatal/panic logfile record for the problem description Apr 18 12:24:31 XXXXgate postfix/qmgr[29035]: warning: transport smtp failure -- see a previous warning/fatal/panic logfile record for the problem description Apr 18 14:21:07 XXXXgate postfix/smtp[2232]: panic: deliver_completed: bad offset 0 Apr 18 14:21:08 XXXXgate postfix/qmgr[29035]: warning: transport smtp failure -- see a previous warning/fatal/panic logfile record for the problem description postfix/smtp[1942]: panic: vstream_buf_alloc: unable to extend fixed-size buffer postfix/master[2915]: warning: process /usr/libexec/postfix/qmgr pid 16054 exit status 1 The result is a performance decrease of mail delivery. I thought I was the only one seeing this problem and, after discussing it on postfix-users, everyone (me included) thought it was a hardware issue. But today someone else emailed me saying that he's having the same problem with a busy FC3/Postfix mail server. The problem does not seem to appear unless the machine is really busy, but please check your servers anyway. Judging from the discussion on postfix-users, it's specific to Fedora, perhaps only to FC3, perhaps only to FC3 with all updates (including the kernel), so maybe it's not a problem with Postfix per se but with something else (kernel? exec-shield? compiler? libraries?). I cannot test it anymore, not too soon anyway. -- Florin Andrei http://florin.myip.org/
