unable to run clusterware package on service guard

[SJ1436]

hello ,
I have Service Guard Cluster installed on HP-UX 11.23 (rp3440)
when trying to run the clusterware package (clstwrpkg) it is giving the following error:

root@ab-dwh1:/# cmrunpkg -v clstwrpkg
Running package clstwrpkg on node ab-dwh1
The package script for clstwrpkg timed out and was terminated via a signal. Manual cleanup
is required
Unable to run package clstwrpkg on node ab-dwh1
Check the syslog and pkg log files for more detailed information
Running package clstwrpkg on node ab-dwh2
The package script for clstwrpkg timed out and was terminated via a signal. Manual cleanup is required
Unable to run package clstwrpkg on node ab-dwh2
Check the syslog and pkg log files for more detailed information
cmrunpkg: Unable to start some package or package instances

anyone has an idea on how to resolve a such issue?
do you have the procedure for the manual clean up?

thanks in advance for your assistance.
kind regards.
SJ

[Steven E. Protter]

Shalom SJ,

Check the syslog and pkg log files for more detailed information

/var/adm/syslog/syslog.log

We need more to help you.

SEP

[Rasheed Tamton]

> Check the syslog and pkg log files for more detailed information

SEP also said it.

The pkg log file will be in /etc/cmcluster/yourPKGname

If you go through both the log (control.sh.log & syslog.log) files, you will certainly get some clues reg. the problem.

Rgds.

[SJ1436]

Hi all,

Thanks for your answers.
I have tried to halt the cluster but I couldn't.
when trying to get a look on the syslog.log, the problem was that "vgcluster_vote volume group is still active", which is the voting disk
what can I do in this case, sure nothing related to this in the BDF command , and nothing related to it in mnttab file.
I tried to stop it manually but i couldn't since it is always "device busy", even the fuser command didn't help.
any idea??
I will attach the package logfile as soon as possible.

thanks and kind regards,
SJ

[SJ1436]

hi again,

Pleas find uploaded a portion of the syslog after restarting the server. there are some warnings.

thanks for your assistance

[SJ1436]

hello,

kindly find uploaded the package log files on both servers of the cluster.
sorry for the mistake , i the portion that I have uploaded in my previous post is from the package log file and the the syslog.

thanks a lot
SJ

[Rasheed Tamton]

What is the status of the below disks:

/dev/dsk/c6t0d4
/dev/dsk/c6t0d5
/dev/dsk/c6t0d2
/dev/dsk/c8t0d3

Does ioscan display all those as claimed.

[SJ1436]

Hi Rasheed,

thanks for your answer.
in fact those disks does not appear in the ioscan command. and after diagnosing the problem I found that one of the FC cards of the server is damaged , I will change it as soon as possible.
but do you think that this can be the cause of the problem ? I mean that this should not cause the package not to run at least on the first node on which everything is good.
what do you think ?
do you have any suggestions?

thanks and kind regards,
SJ

[Rasheed Tamton]

Your package was not started after Dec. 9. So, I assume it is a test/dev server or the pkg is such.

As you did not specify the setup of the pkg, I can't surely say. Is it active/active or active/passive.

Can you do a vgdisplay on all these VGs from both the nodes:

vgdisplay vgclwr_vote
/dev/vgclwr_vote
/dev/vgclwr_ocr
/dev/vgora1_asm
/dev/vgora2_asm
/dev/vgora3_asm
/dev/vgora4_asm

If you can't see all then may be you have to do all the VGs cluster aware:

vgchange -c y /dev/vgora4_asm (on all VGs)
vgchange -a y or -a e on all VGs

As my weekend starts here, I am afraid I can't follow up the thread until SAT. Hope someone will pop up and help you if the vgchange command does not work.

Regards.

[SJ1436]

Hi Rasheed,

Thanks again for your help.
but we can always run the vgdisplay for all the listed VG on both servers.
do you think it gonna be useful to deactivate them and then reactivate ?

another thing, I have changed the deffective FC card , and it is working normally. but the service still cant find the alternate paths. do I need to reconfigure anything after the Hardware replacement?

Thanks again.
SJ

[Torsten.]

"I have changed the defective FC card , ... do I need to reconfigure anything after the Hardware replacement?"

Do you use WWN based zoning? If you do, you need to adjust the config on the switch.

[Rasheed Tamton]

>I have changed the deffective FC card
Is this FC claimed by ioscan. Otherwise, you have to do a insf -e

You have to deactivate the volume groups for the clusters and make them cluster aware as in my above post and enable it again.

[SJ1436]

Hi Rasheed,

Thanks again.
in fact I have searched the ioscan and found 2 entires for the "fcparray" which corresponds to one card I think (I have 2 cards on each server). on the other server there are 4 entries.
I have made insf -e but nothing changed, I have then restarted the server but it did not help.
what do you think?

Best Regards,
SJ

[SJ1436]

Hello ,

could anyone advise how can I access the initiator settings to check them ??
thanks
SJ

[Rasheed Tamton]

ioscan -fnCctl (look for the initiator)
ioscan -fnCfc

If you have card which is not found by the system then one reason is that it is not seated correctly on the system. You have to inspect the system and see how many of them are there on the system.

[Torsten.]

Could you post the output of "ioscan -fn" please?

Did you already check the zoning configured on your switches?

Maybe you need to adjust the WWN based presentation on your array too.

[SJ1436]

Hi,

Thanks for your answers.
how can I check and edit the "Initiator Type" settings ?

Best regards

[Torsten.]

For FC cards the HBA will sense the settings with the switch - you can't change them on the host.
Check with (example):

# fcmsutil /dev/fcd0

(or whatever applies).

[Rasheed Tamton]

It is better to give the output of ioscan -fn as asked by Torsten. That would give a good idea about the hw in your system.

ioscan -fnCfc
will show the dev path for the FCs and do fcmsutil on it.

[SJ1436]

Hello,

Thanks again.
I did the ioscan and the fc card is claimed .
the fcmsutil just give info about the card, i need to check the initiator type and change it if required. could you please advice ?

Thanks you
SJ

[SJ1436]

Hello,

Thanks again.
I did the ioscan and the fc card is claimed .
the fcmsutil just give info about the card, i need to check the initiator type and change it if required. could you please advice ?

Thank you
SJ

[Torsten.]

If you hide all the requested information, you will be alone, I guess.

[SJ1436]

hi all ,

Please find attached the required info.
It is not that I want to hide info , but it is a server at a client and he has some restriction on getting away some information on the system.

Thanks a lot
SJ

[Torsten.]

a previous question was:

"What is the status of the below disks:

/dev/dsk/c6t0d4
/dev/dsk/c6t0d5
/dev/dsk/c6t0d2
/dev/dsk/c8t0d3

Does ioscan display all those as claimed."

All the EVA vdisks are c10... and c11... related to hardware path 0/3/1/.

The card on 0/4/1 does not show devices.


Again, if you use WWN based zoning on the switch - adjust it to the new WWN of the HBA.



The EVA presents vdisks to the servers based on their HBA WWN - you need to adjust the EVA config to present vdisks to this new card.

N_Port Port World Wide Name = 0x50060b000012f0c6

This WWN must be used on the EVA presentation settings!

[SJ1436]

Hi Torsten,

Thanks for your reply.
In fact , this is the problem that there are disks that are not shown by the new added card.
I have already adjusted the zoning configuration , even I deleted the zone and recreate it again , and add it the the zone config and restarted the server, and I always have the same problem.
and since the card is claimed , that's mean that it is working.
anything else can I check ?

Best regards,
SJ