dmesg

[RUSSOCKI Wladyslaw]

the dmesg command tells me "The diagnostic logging facility has started receiveing excessive errors from the I/O subsystem" I/O error will be lost until the cause of the excessive I/O logging is corrected...

this message appears at boot time but rarely after..

STM shows me that the system is all right an no errors appears...i check that with the hard support.

Has someone already seen that?
thanks'

[Rick Garland]

dmesg does not provide a date-timestamp with the output. How does the syslog look?
You will have date-timestamps with those messages and can see if the messages are older.

[Vince Inman]

I've seen this happen quite often. It indicates that the diagnotics daemon is receiving too many errors and that it can't keep up with the errors being reported.

syslog may have some messages related to the problem. as well you may want to peruse the
/var/stm/logs/sys/activity_log for any messages that may be indicative of a h/w problem.

Beforehand I would stop and start diagmond and see if the problem disappears.

[Kofi ARTHIABAH]

I have had that situation before and it was a failing device - it was not completely dead, but it was dieing. You might want to check /var/log/syslog/syslog.log and /var/log/syslog/OLDsyslog.log for clues... because if at the time you run stm, there is nothing wrong (ie. problem is intermittent) then you cannot catch it with STM, you will have to check the syslogs for more clues.

You can also check the root mail for any outputs.

[John Palmer]

Yes I've seen the message but only when there were genuine errors, not specifically at boot time.

STM should indicate what is causing the errors but it sounds as though you've already been down that route.

When you say 'boot time' is this when you power on the server or just when you reboot? I was wondering whether some hardware was possibly taking a little while to 'settle down' after powering up.

[Vince Inman]

John:

You are correct, I stand corrected.

Vince

[Victor BERRIDGE]

As Rick mentionned, dmesg is not reliable since its based on reading /var/adm/syslog/syslog.log

The same appears when you have a file system full and you kown you cleared the file system or extended it...
To have reliable messages using dmesg would be to reinitialize the syslog.log mail.log and then /sbin/init.d/syslogd stop / start

[RUSSOCKI Wladyslaw]

thanks for all theses replies!
exact : I find the same message in the /var/adm/syslog/syslog.log file, but it does'nt explain to me what is the origin of the excessive IO errors. Given that for STM all my devices are ok, that's very strange.
I've stopped/starte the diagmon daemon, and nothing happens! that is to say that, at this time the message does'nt appear in the syslog.log file.
Is someone kwowing this pb and using FC Mux?
thanks'
M

[Rick Garland]

Part of querying the device is to see that all functions are performing, this would include the error functions. The device is probably "settling down" after the initial queries.

[John Palmer]

HP have had a lot of problems with Fibre Channel controller cards.

We had problems which only manifested themselves during periods of very heavy disk I/O.

Since we've had the cards updated though all the problems have gone away.

[Kofi ARTHIABAH]

When I had that problem, it was indeed with my FC (fibre channel adapter) HP support replaced fibre channel card. He mentioned that HP found that a series of the Fibre cards had a problem with their laser modules (GLM) and typically started failing after about 9 months... solution? get HP to replace your fibre channel card.

BTW is your machine a D370 by any chance?

[Tom Danzig]

As stated, it's probably a bad fibre channel card. Check /var/adm/syslog/syslog.log for messages to that effect. If one of the cards is bad, you'll see thousands of messages flooding the syslog.