[jadmin] jabberd2 + debian, frequent client disconnects

[henry mason]

	greetings jabber-admin;

	i'm having this problem that just surfaced within the last
	month or so, *possibly* after an upgrade to jabberd2-mysql
         2.0s10-1, but not positive.

	this only happens on high latency connections, in this
	particular case it's satellite connections that are really
	being affected. clients that connect through a T1, OC or
	DSL connection appear to be fine.

	i have /usr/sbin/jabberd2-sm -D redirected to a file, and
	have been trying to dig through the output to find any
	clues.

	it seems to just happen randomly:

Fri Jun 30 10:25:25 2006 c2s.c:97 read 1 bytes
sx (io.c:181) passed 1 read bytes
sx (chain.c:93) calling io read chain
sx (io.c:205) decoded read data (1 bytes):
Fri Jun 30 10:25:25 2006 c2s.c:381 read action on fd 9
sx (io.c:159) 9 ready for reading
sx (io.c:165) tag 9 event 2 data 0x9b358
Fri Jun 30 10:25:25 2006 c2s.c:43 reading from 9
Fri Jun 30 10:25:25 2006 c2s.c:97 read 1 bytes
sx (io.c:181) passed 1 read bytes
sx (chain.c:93) calling io read chain
sx (io.c:205) decoded read data (1 bytes):
Fri Jun 30 10:25:25 2006 c2s.c:381 read action on fd 11
sx (io.c:159) 11 ready for reading
sx (io.c:165) tag 11 event 2 data 0x15a438
Fri Jun 30 10:25:25 2006 c2s.c:43 reading from 11
Fri Jun 30 10:25:25 2006 c2s.c:97 read 1 bytes
sx (io.c:181) passed 1 read bytes
sx (chain.c:93) calling io read chain
sx (io.c:205) decoded read data (1 bytes):
Fri Jun 30 10:25:25 2006 c2s.c:381 read action on fd 25
sx (io.c:159) 25 ready for reading
sx (io.c:165) tag 25 event 2 data 0x9b358
Fri Jun 30 10:25:25 2006 c2s.c:43 reading from 25
Fri Jun 30 10:25:25 2006 c2s.c:97 read 1 bytes
sx (io.c:181) passed 1 read bytes
sx (chain.c:93) calling io read chain
sx (io.c:205) decoded read data (1 bytes):
Fri Jun 30 10:25:25 2006 c2s.c:381 read action on fd 9
sx (io.c:461) 9 state change from 4 to 6
sx (io.c:462) tag 9 event 7 data 0x0
Fri Jun 30 10:25:25 2006 c2s.c:400 close action on fd 9
Fri Jun 30 10:25:25 2006 [notice] [9] [67.143.48.202, port=6268] disconnect
Fri Jun 30 10:25:25 2006 sm.c:61 built new route nad for x30back at jabber.dbsinet.com/Gaim action end target (null) id (null)
sx (chain.c:106) calling nad write chain
sx (io.c:369) queueing for write: <route xmlns='http://jabberd.jabberstudio.org/ns/component/1.0' from='c2s' to='jabber.dbsinet.com'><sc$
sx (io.c:392) tag 5 event 1 data 0x0
Fri Jun 30 10:25:25 2006 c2s.c:549 want write
Fri Jun 30 10:25:25 2006 c2s.c:1060 write action on fd 5
sx (io.c:293) 5 ready for writing
sx (io.c:251) encoding 251 bytes for writing: <route xmlns='http://jabberd.jabberstudio.org/ns/component/1.0' from='c2s' to='jabber.dbsi$
sx (chain.c:79) calling io write chain
sx (io.c:312) handing app 251 bytes to write
sx (io.c:313) tag 5 event 3 data 0x9b358
Fri Jun 30 10:25:25 2006 c2s.c:586 writing to 5
Fri Jun 30 10:25:25 2006 c2s.c:590 251 bytes written
sx (io.c:346) tag 5 event 0 data 0x0
Fri Jun 30 10:25:25 2006 c2s.c:544 want read
sx (io.c:394) tag 5 event 0 data 0x0
Fri Jun 30 10:25:25 2006 c2s.c:544 want read
Fri Jun 30 10:25:25 2006 c2s.c:381 read action on fd 11
sx (io.c:461) 11 state change from 4 to 6
sx (io.c:462) tag 11 event 7 data 0x0
Fri Jun 30 10:25:25 2006 c2s.c:400 close action on fd 11
Fri Jun 30 10:25:25 2006 [notice] [11] [67.143.211.218, port=3220] disconnect
Fri Jun 30 10:25:25 2006 sm.c:61 built new route nad for x34back at jabber.dbsinet.com/Gaim action end target (null) id (null)
sx (chain.c:106) calling nad write chain

	note how both connections were reading fine, then suddenly
	had a state change, forcing them to disconnect.

	i'm not sure if this actually indicates a disruption in the
	connection, or not. supposedly the connections don't go down
	during these disconnects (according to the people using them,
	but...)

	clients are using a mixture of Gaim and Psi. Psi is configured
	to send keepalive packets to the server. same problems.

	any ideas?

	thanks,
	henry