session manager don't reopen log at kill -HUP

[jens]

I using file logging and rotate the logfiles with logrotate. On the end of the logrotate script I will reopen log with kill -HUP cat /var/run/jabber/sm.pid Several times work that, but than the session manager don't reopen the log files. The other components, don't have this problem.

And: From the first kill -HUP, jabber:iq:last don't work.

[jens]

Why nobody reads my task?

[smoku]

We do read your ticket. It's just that nobody took up the problem yet.

[shax]

I have same issue. OS type: OpenBSD 4.2 amd64 log type = 'file' When i send HUP signal to sm:

SM  : Mon Jun  2 21:39:00 2008 [notice] HUP handled. reloading modules...
ROUT: sx (chain.c:106) calling nad write chain
SM  : sx (io.c:196) tag 5 event 2 data 0x4fa2de00
S2S : sx (io.c:196) tag 5 event 2 data 0x4e5bc240
C2S : sx (io.c:196) tag 5 event 2 data 0x49db1740
ROUT: Mon Jun  2 21:39:00 2008 [notice] [127.0.0.1, port=48759] disconnect
SM  : Mon Jun  2 21:39:00 2008 [notice] modules search path: /usr/local/lib/jabberd
ROUT: sx (io.c:402) queueing for write: <presence xmlns='http://jabberd.jabberstudio.org/ns/component/1.0' from='s2s'/>
SM  : Mon Jun  2 21:38:37 2008 sm.c:53 reading from 5
S2S : Mon Jun  2 21:38:37 2008 router.c:44 reading from 5
C2S : Mon Jun  2 21:38:37 2008 c2s.c:669 reading from 5
SM  : Mon Jun  2 21:39:00 2008 [notice] module 'status' added to chain 'sess-start' (order 0 index 0 seq 0)
ROUT: sx (io.c:425) tag 7 event 1 data 0x0
SM  : Mon Jun  2 21:38:37 2008 sm.c:78 read 154 bytes
S2S : Mon Jun  2 21:38:37 2008 router.c:69 read 154 bytes
C2S : Mon Jun  2 21:38:37 2008 c2s.c:694 read 154 bytes
SM  : Mon Jun  2 21:39:00 2008 [notice] module 'status' added to chain 'sess-end' (order 0 index 0 seq 1)
ROUT: Mon Jun  2 21:38:37 2008 router.c:521 want write
SM  : sx (io.c:212) passed 154 read bytes
S2S : sx (io.c:212) passed 154 read bytes
C2S : sx (io.c:212) passed 154 read bytes
SM  : Mon Jun  2 21:39:00 2008 [notice] module 'iq-last' added to chain 'sess-end' (order 1 index 1 seq 0)
ROUT: Mon Jun  2 21:38:37 2008 router.c:872 write action on fd 7
SM  : sx (chain.c:93) calling io read chain
S2S : sx (chain.c:93) calling io read chain
C2S : sx (chain.c:93) calling io read chain
SM  : Mon Jun  2 21:39:00 2008 [notice] module 'validate' added to chain 'in-sess' (order 0 index 2 seq 0)
ROUT: sx (io.c:324) 7 ready for writing
SM  : sx (ssl.c:380) in _sx_ssl_rio
S2S : sx (ssl.c:380) in _sx_ssl_rio
C2S : sx (ssl.c:380) in _sx_ssl_rio
SM  : Mon Jun  2 21:39:00 2008 [notice] module 'status' added to chain 'in-sess' (order 1 index 0 seq 2)
ROUT: sx (io.c:282) encoding 79 bytes for writing: <presence xmlns='http://jabberd.jabberstudio.org/ns/component/1.0' from='s2s'/>
SM  : sx (ssl.c:384) loading 154 bytes into ssl read buffer
S2S : sx (ssl.c:384) loading 154 bytes into ssl read buffer
C2S : sx (ssl.c:384) loading 154 bytes into ssl read buffer
ERROR: sm died.  Shutting down server.
JBRD: Got a signal... pass it on.
JBRD: It was a TERM.  Shut it all down!

[shax]

sorry... i forgot jabberd version: 2.2.0 (2.1.24.1 also)

[lavda]

It seems that ever since dynamic reloading of modules was integrated as part of the logrotate (in sm), sm crash if there are current active sessions or sessions starting/stopping at the moment you run '/bin/kill -HUP cat /usr/local/var/jabberd/pid/sm.pid 2> /dev/null. If I look at the core dump, it segfault at different places in the code but always because some address is no longer valid.