On Mon, 2008-08-25 at 18:39 +0300, Lauri Jesmin wrote:
Hello,
I upgraded dovecot to 1.1.2 this weekend and i get those errors: grep dovecot /var/log/net/maillog |grep Panic | cut -d " " -f 11-100| sort | uniq -c | sort -rn 7547 file index-sync.c: line 39: assertion failed: (seq_range_exists(&ibox->recent_flags, uid))
This should help: http://hg.dovecot.org/dovecot-1.1/rev/8cc0eaec7d0f
89 file index-mail.c: line 1091: assertion failed: (!mail->data.destroying_stream)
I've seen this before, but I've really no idea how to reproduce it. Can you get a gdb backtrace from this? See http://dovecot.org/bugreport.html
18 file mbox-sync-rewrite.c: line 590: assertion failed:(mails[idx].from_offset == start_offset)
Either because of an internal bug or because the mbox file changed unexpectedly. Possibly due to NFS read cache issues.
5 file ostream-file.c: line 533: assertion failed: ((size_t)ret ==total_size || !fstream->file)
This should help: http://hg.dovecot.org/dovecot-1.1/rev/f690d84a9fcb
5 file mbox-lock.c: line 675: assertion failed: (lock_type ==F_RDLCK || mbox->mbox_lock_type != F_RDLCK)
I'm not sure about this. Can you get gdb backtrace?
1 file message-parser.c: line 684: assertion failed:(ctx->part->physical_pos >= ctx->input->v_offset)
This should help: http://hg.dovecot.org/dovecot-1.1/rev/1d8bd4bc3038
My setup is this: 3 servers, all Solaris, running dovecot, share home directiories via NFS, mbox format mailboxes in home directories. Indexes are separate in every server, not shared via NFS. About 20 000 users total.
What do you use as NFS server? Does it support nano/microsecond precision in mtimes? Could you post your dovecot -n output? You should use mail_nfs_storage=yes at least.