Connection refused: Child 16061failed to establish umbilical to moderator!

Hello!

We are having currently hard Problems with mod_backhand.

Our Environment:

Servers: 	4 Nodes (A,B,C,D), each Dual P3 1Ghz with 1GB Ram.
Software:	apache-ssl 1.3.26 (debian)

The nodes building 2 Groups:
*) lgD
	-) Primary Server Node A
		-) Backhand Server Node C
		-) Backhand Server Node D
*) lgE	
	-) Primary Server Node B
		-) Backhand Server Node C
		-) Backhand Server Node D

Node A & B are Load Balanced through Round-Robin-DNS.

The only one who's makeing troubles is Node B. One or two Weeks ago
backhand starts dieing with the message:

[Thu Jan  1 19:32:14 2004] [error] (111)Connection refused: Child 16061
failed to establish umbilical to moderator!
[Thu Jan  1 19:32:14 2004] [error] (111)Connection refused: Child 16063
failed to establish umbilical to moderator!
[Thu Jan  1 19:32:14 2004] [error] (111)Connection refused: Child 16062
failed to establish umbilical to moderator!
[Thu Jan  1 19:32:16 2004] [error] (111)Connection refused: Child 16064
failed to establish umbilical to moderator!
[Thu Jan  1 19:32:17 2004] [error] (111)Connection refused: Child 16065
failed to establish umbilical to moderator!
[Thu Jan  1 19:32:17 2004] [error] (111)Connection refused: Child 16066
failed to establish umbilical to moderator!
[Thu Jan  1 19:32:18 2004] [error] (111)Connection refused: Child 16071
failed to establish umbilical to moderator!
[Thu Jan  1 19:32:18 2004] [error] (111)Connection refused: Child 16072
failed to establish umbilical to moderator!

So i installed a cronjobs which restarts apache-ssl on Node B to a
specific time during night.

I looked out in this mailing list, found some entries to change
kernel-values in /proc (not helping), looking for apache segfaults in
the errorlogs (not found).

only found two line in the error-log before the crash:

[Thu Jan  1 19:06:02 2004] [error] (111)Connection refused: Child 11432
failed to establish umbilical to moderator!
[Thu Jan  1 19:06:02 2004] [notice] child pid 29408 exit signal Broken
pipe (13)

first the backhand-error-message, then an apache error.



Now we have the Problem, that backhand crash much more often then one
times a day - in different time intervals.

there where no changes on the hardware, also not on the software
(apache, php, mod_perl, websites and so on...).

I checked all, /var/backhand looks good. Checked the PID of the
Moderator.... all ok....


Does anyone has an idea howto solve this Problem?


Andreas

CUBiT Information Technology