cluster shutdown problem

[Sajjad Sahir]

I shutdown the cluster by using cmhalt -f
when i am using the cmviewcl it is showing all package unowned.

after some time I want to start the cluster again but it saying cluster is still running on hp_data1 and hp_data2
why cluster not shutdown properly with the cmhalt -f command?

my cluster name is hp_data inside these I have two nodes are there hp_data1 and hp_data2
4 packages are running orapkg, nfs1 and orpkg2 and nfs2

[Steven E. Protter]

Shalom,

Check /var/log/messages
Check any cluster logs set up in your cluster configuration and package configuration. There will be clues.

Common problems:
Oracle's dbshut command does a shutdown on the database. This waits for all records to be closed and then shuts the database. It needs to be modified to do a shutdown immediate

NFS hangs a lot if there are open connections.

The answer to this mystery is in the logs.

SEP

[skt_skt]

"I shutdown the cluster by using cmhalt -f
when i am using the cmviewcl it is showing all package unowned"

why do you want to shutdown the cluster forcefully?is the normal one NOT wokring?is there is any related problem due to which u r halting the cluster? did u verify the cluster was down?

no command called "cmhalt"

ll cmhalt*
-r-xr-xr-x 11 bin bin 3136024 Sep 29 2005 cmhaltcl
-r-xr-xr-x 11 bin bin 3136024 Sep 29 2005 cmhaltnode
-r-xr--r-- 10 bin bin 3136024 Sep 29 2005 cmhaltpkg
-r-xr--r-- 10 bin bin 3136024 Sep 29 2005 cmhaltserv

which one did u try?


you cant halt a cluster(cmhaltcl) if there is active nodes.

shutdown packages(cmhaltpkg) and then node cmhaltnode and then go to cmhaltcl

could u post "cmviewcl" o/p

[Asif Sharif]

Salam Sajjad,

Why you are going to shutdown cluster forcefully, for some reason if you want to do then use this command.
#cmhaltcl -v -f

Check syslog for more detail "/var/adm/syslog/syslog.log".


Regards,
Asif Sharif

[Stephen Doud]

Cluster actions are recorded in the system log.
As stated before, refer to that log, whether it be /var/adm/syslog/syslog.log (or /var/log/messages if you meant to log this thread in the Linux->Serviceguard forum).

In the system log, you should find diagnostic messages about Serviceguard's activities. Serviceguard uses several daemons to accomplish it's work, so look for messages that contain cmcld or cmclconfd or other daemons beginning with the letters cm.

When you locate problematic messages, report them here, or, if the problem is urgent, consider opening a software support case with the HP Support Center.

[Sajjad Sahir]

dear asif I used cmahltcl -f
command after some time I used cmruncl -f
that time i am gettig the message cluster is still running on my both nodes.
I used cmviewcl to see the out put

all packages are unowend but the problem is
cluster is still running on both nodes

[skt_skt]

check with "vgdisplay" if any cluster(exclusive) VG is still active.Inthat case cluster wont be stopped.

[Anshumali]

"I used cmahltcl -f
command after some time I used cmruncl -f"

Can you post your cmviewcl o/p?

Looks like your packages are unowned. What is the state of PKGs in cmviewcl?
If they are down, you may need to do cmmodpkg -e -n

Post it..

[Sajjad Sahir]

pacakges are unowend all 4 packages unowned

but cluster is not shutting down properly

[Sajjad Sahir]

dear anshumali
cmviewcl showing all packages unowned
the it display the message cluster is still running on node1 and node2 while I am using cmruncl
that means it was not shutdown properly

[Rita C Workman · Accepted solution]

Sajjad,

Read the responses above....

Probably you have at least one package that did not go down right. Look in your package.cntl.log file to see. You can also run bdf on your hosts to see if any package mountpoints are still showing. If they are you need to umount it, then vgchange -a n then vgchange -c y .

Definitely you would have cluster errors reporting in your syslog file as others pointed out. Go back, find the errors and address or at least post them here so we'll know what the exact problem was and then someone can help you.

Rgrds,
Rita

[Mridul Shrivastava]

Please check the things suggested by the guys above like syslog.log, cntrl.log and post o/p of cmviewcl.
You can check the following daemon is running or not

ps -ef |grep cmcld

If this daemon is running then cluster did not halt properly. To know the reason why it didn't halt properly we need to see those logs, mentioned above.

[Sajjad Sahir]

closed