[985] in Zephyr_Bugs
Zhm loops on HP_UX 10.20 host
daemon@ATHENA.MIT.EDU (Peter-Klaus Schilling)
Thu Dec 23 09:55:19 1999
Date: Thu, 23 Dec 1999 15:54:23 +0100 (MET)
From: Peter-Klaus Schilling <pks@mail.desy.de>
Reply-To: Peter-Klaus Schilling <peter-klaus.schilling@desy.de>
To: bug-zephyr@MIT.EDU
Cc: ghudson@MIT.EDU
Message-Id: <Pine.HPX.4.10.9912231519260.21943-100000@puls05.desy.de>
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Some years ago we installed the Zephyr system on a heterogenous network of
Unix hosts connected via the AFS file system. So far everything worked
fine. But now we have trouble with zhm on one HP_UX host:
We have a number of hosts using HP_UX 10.20 of a certain patch level and
everything -- especially zhm, which is started by inetd -- works fine
(allthough the binaries have been generated under HP_UX 9.x.). Recently we
prepared a new HP_UX 10.20 (more patches), which is to be used on all
HP-hosts in the near future.
And now, lo and behold, zhm loops, using nearly all available CPU-time,
when started by inetd on this new system. It behaves reasonably, when
started manully by root! And: recompilation on the new host doesn't help.
On the other hand, this "looping" zhm doesn't show any operational
deficiencies on this host.
I recompiled zhm using the -DDEBUG option, started it manually under root
with the -d option and redirected stderr to a file:
/usr/local/bin/sbin/zhm -d zephyr1.desy.de 2>/tmp/zhm.debug &
As expected it worked fine, didn't loop, and wrote debugging information
to the file, which starts like this:
Queue initialized and flushed.
Server = wotan.desy.de
zephyr server port: 2103
zephyr client port: 2104
Waiting for a packet...Got a packet.
notice:
z_kind: 5
z_port: 2104
z_class: ZEPHYR_CTL
z_class_inst: HM
z_opcode: BOOT
z_sender: HM
z_recip:
z_def_format:
z_message: SENT
A notice came in from the server.
Retransmitting queue to new server...
Waiting for a packet...Waiting for a packet...Notice timeout
Resending notices...
Waiting for a packet...Notice timeout
Resending notices...
Now, when I have inetd start zhm like this:
zephyr-hm dgram udp wait root /usr/local/sbin/zhm zhm -d -i
zephyr1.desy.de 2>/tmp/zhm.debug
(one line in /etc/inetd.conf)
it won't even touch the file /tmp/zhm.debug!
I'm completely lost now! I would at least expect the first debugging line
"Queue initialized and flushed." to show up, since as far as I see from
the c-source nothing dramatic happens up to this point.
Could you please give me some advice, how to proceed?
Best regards,
Peter Schilling
--------------------------------------------------
e-mail: peter-klaus.schilling@desy.de
address: Peter K. Schilling
DESY Dpt. IT
Notkestr. 85 - D-22607 Hamburg - Germany
phone: +49 40 89983638 fax: +49 40 89943638
--------------------------------------------------
