On Thu, 2006-08-10 at 18:44 -0500, Jeremy C. Reed wrote:
On Thu, 2006-08-10 at 17:40 -0500, Jeremy C. Reed wrote:
Was running dovecot-1.0.rc2 but many stale imap processes (only one user of IMAP).
Switched to dovecot-1.0.rc6 but problem continued.
I don't really understand this problem.. Is this user using SSL? What if you try the attached patch, does it change anything?
No ssl as far as I know. (I tried to explain in my previous email: no imaps in protocols and not using stunnel for imaps for over week. Also the end user said stopped using imaps over a week ago to make sure that was not the problem.)
Oh. I thought you meant you used Dovecot SSL instead of stunnel to make sure the problem wasn't with it, or something. :)
Could you get process traces and gdb backtraces from them? See http://dovecot.org/bugreport.html#hangs
Attaching to program: /local/service/dovecot/libexec/dovecot/imap, process 49042 Reading symbols from /usr/lib/libc.so.4...done. Reading symbols from /usr/libexec/ld-elf.so.1...done. 0x2811d590 in kevent () from /usr/lib/libc.so.4 (gdb) bt full #0 0x2811d590 in kevent () from /usr/lib/libc.so.4 No symbol table info available. #1 0x80a8b20 in event_callback (context=0x80d8030) at ioloop-notify-kqueue.c:41 ctx = (struct ioloop_notify_handler_context *) 0x80d8030 io = (struct io *) 0x16b ev = {ident = 135045792, filter = 1, flags = 0, fflags = 134910204, data = 135098368, udata = 0x80dd050}
So it's at least not in some infinite loop. What about ktrace, does it show anything else than kevents?
Also, what do these commands say after attaching to gdb:
p *my_client p *(struct file_istream *)my_client.input.real_stream p *(struct file_ostream *)my_client.output.real_stream
And if I understood correctly, the same hangs happened even if you didn't use --with-ioloop parameter at all, ie. it used poll?