I use Cyrus 2.1.9. Sometimes ctl_deliver hangs until I restart cyrus.
Looks like it waits for unlocking of deliver.db. That's what ktrace shows:
54735 ctl_deliver CALL pread(0x5,0x282b47d4,0x2000,0,0x27e000,0)
54735 ctl_deliver GIO fd 5 read 8192 bytes
[....skip....] [EMAIL PROTECTED]>\0user.atbmarket\0000\^D\0\^Argэ>a+\0\^A<618DCCJLLF3C6H57\
@translit.ru>\0user.servag\0da\^D\0\^A\^Srш>1H\0\^A<6186081.10545702618\
[EMAIL PROTECTED]>\0user.loveme\0J"
54735 ctl_deliver RET pread 8192/0x2000
54735 ctl_deliver CALL select(0,0,0,0,0xbfbff854)
54735 ctl_deliver RET select 0
54735 ctl_deliver CALL select(0,0,0,0,0xbfbff854)
54735 ctl_deliver RET select 0
54735 ctl_deliver CALL select(0,0,0,0,0xbfbff854)
54735 ctl_deliver RET select 0
54735 ctl_deliver CALL select(0,0,0,0,0xbfbff854)
54735 ctl_deliver RET select 0
54735 ctl_deliver CALL select(0,0,0,0,0xbfbff854)
54735 ctl_deliver RET select 0
54735 ctl_deliver CALL select(0,0,0,0,0xbfbff854)
54735 ctl_deliver RET select 0
and select waits until kill of program.
I use db3_nosync as deliver backend. Version: Sleepycat Software: Berkeley DB 3.3.11: (July 12, 2001)
Can I avoid such hangs?
-- Best wishes, Dmitry Sergienko
