I have an HACMP configured which has two nodes participating in
the cluster.The problem here is the Group Services is continously
logging the error Saying that its not able to write into the log file.I
will paste the screen shot of the error below.
So at first i tried recreating the log file and recycled the
syslogd ,but it was very odd that i found the log file was deleted when
i looked from the log directory.The same thing happened for next two
tries as well. The HACMP was also restarted twice(The server was
rebooted even though not as a part of solving this ) , but the error
is still getting logged in the error log.
The error log looks like below :


Description
Group Services informational message

Probable Causes
Informational message

Failure Causes
Informational message

Recommended Actions
NONE

Detail Data
DETECTING MODULE
RSCT,TraceStream.C,1.79,678
ERROR ID
60.oOd0Ps0j1/KVX.5JV08....................
REFERENCE CODE

DIAGNOSTIC EXPLANATION
ERROR writing to log file /var/ha/log/grpsvcs_2_27.ABSAP (rdstate=4
errno=9[A file descriptor does n


I will be thankfull if you can provide a solution
for this problem.

praveen wrote:
> DIAGNOSTIC EXPLANATION
> ERROR writing to log file /var/ha/log/grpsvcs_2_27.ABSAP (rdstate=4
> errno=9[A file descriptor does n

Hm, what does the file descriptor not? Please cite the full
error log message. Meanwhile you could take a look at the
cluster log files (hacmp.out, cluster.log) for something
unusual. Seems something/someone's deleting your logfile.
Does it stay there if you don't refresh syslogd? How do
you rotate your logfiles?

Regards,

Frank

praveen wrote:
....
>
> DIAGNOSTIC EXPLANATION
> ERROR writing to log file /var/ha/log/grpsvcs_2_27.ABSAP (rdstate=4
> errno=9[A file descriptor does n
>
>
> I will be thankfull if you can provide a solution
> for this problem.
>

Praveen,

I have similar messages in my HACMP 5.1 on AIX 5.2.6 cluster.
Seems harmless though, the cluster is functioning very well.
Probably a bug in grpsvcs on 5.2

"
ERROR writing to log file /var/ha/log/grpsvcs_3_16.aixcl.long (rdstate=6
errno=3[The process does no
"

Gert

Hi Frank,

The error log also gives only that much. Hopefully it should be
"descriptor does not exist".
Could not see any thing wrong in hacmp log file (hacmp.out) .
And there is no application that deletes Even i was surprised to see
the files getting deleted.
And as you have suggested i too feel the culprit is the log rpotater,
could to give some tips to check how its configured. It was configured
by person before me and i am a baby in HACMP

Hi Gert

Yep yours also seems to be the same problem. Did you try recreating the
log file if so then is it geting deleted?

cheers

praveen wrote:
> Hi Gert
>
> Yep yours also seems to be the same problem. Did you try recreating the
> log file if so then is it geting deleted?
>
> cheers
>
In my case the log is still there and the problem seems to be limited to
not being able to write 1 line:
"
DIAGNOSTIC EXPLANATION
Lost 1 lines to log file /var/ha/log/grpsvcs_3_16.aixcl. Writing to log
now.
"

It only happens at a cluster start. Something related to restart of a
daemon or a temp lock on the logfile I think.

Gert

praveen wrote:
> The error log also gives only that much. Hopefully it should be
> "descriptor does not exist".
> Could not see any thing wrong in hacmp log file (hacmp.out) .
> And there is no application that deletes Even i was surprised to see
> the files getting deleted.
> And as you have suggested i too feel the culprit is the log rpotater,
> could to give some tips to check how its configured. It was configured
> by person before me and i am a baby in HACMP

Unfortunately logrotate doesn't come with AIX out of the box.
You'll have to look at the usual places (/etc, /usr/local/etc,
....) if you find a logrotate.conf. Might also be some custom
made script doing logfile rotation.
Though gert suggested in his post, that it might be a known
issue with HACMP. What version are you running? Maybe this
has been already fixed with a maintainence release. Still
i'd open a case with IBM to address this - even if it's just
for incrementing the complaint counter for this issue ;-)

Regards,

Frank