1748205 Members
4455 Online
108759 Solutions
New Discussion юеВ

Quorum disk removal

 
SOLVED
Go to solution
ani007
Super Advisor

Quorum disk removal

Hi,
i want to remove the quorum disk from one of our Linux box (Enterprise Linux Enterprise Linux Server release 5.1).I am attaching the current setup.Please find the attachment.Could you please let me know after stopping the pkg & cluster what i have to do .or what will be the procedure.
9 REPLIES 9
Steven E. Protter
Exalted Contributor

Re: Quorum disk removal

Shalom,

Note that RHEL 5.1 was not the most stable of releases. Actual functionality may vary. Going to 5, update 4 might get you better results.

If there is no quorum disk, the cluster might continue to function with fencing. Quorum disk is not a requirement.

If there is no valid fencing mechanism in place, your cluster would go off line and you might need to establish a new quorum disk before your application runs.

What is your fencing mechanism?

SEP
Steven E Protter
Owner of ISN Corporation
http://isnamerica.com
http://hpuxconsulting.com
Sponsor: http://hpux.ws
Twitter: http://twitter.com/hpuxlinux
Founder http://newdatacloud.com
ani007
Super Advisor

Re: Quorum disk removal

As i am new in linux please let me know what information i have to provide.
Matti_Kurkela
Honored Contributor
Solution

Re: Quorum disk removal

Showing your /etc/cluster/cluster.conf file would have been the easiest way to show all the necessary details.

Of course you should remove/obscure any IP addresses and fencing logins/passwords before publishing it.

Stopping the cluster is not necessary.

On any node, while the cluster is running, copy the current cluster.conf to a temporary location:

cp /etc/cluster/cluster.conf /tmp/cluster.conf

Then edit the /tmp/cluster.conf file:
- increment the cluster config_version
- as you have a two-node cluster, set the cman two_node parameter to 1 (=true)
- set the cman expected_votes parameter to 1
- remove the entire block (i.e. all lines from line "", inclusive)

Then run:

ccs_tool update /tmp/cluster.conf

This will update the cluster configuration on all cluster nodes.

After the update is successful, your qluster will no longer be using the quorum disk.
Then you can run

service qdiskd stop
chkconfig qdiskd off

on each node, to shutdown the now-unused qdiskd service.

MK
MK
ani007
Super Advisor

Re: Quorum disk removal

Thank you Matti.
1.We have the outage window.

2.It's a two node cluster.( tneld53n.adr.deep.com & tneld52n.adr.deep.com)

3.I have attached the o/p of clustat & cluster.conf file of one server.

4.please correct me if i am wrong.
First i will make the changes on 52n.After making the changes to cluster.conf file i will reboot it & if its up then i will copy the cluster.conf file to the other node (53n).

currently 2 services are running on 52n

service:ebrdo41d tneld52n.adr.deep.com started
service:ecrde0q tneld52n.adr.deep.com started

So i will stop the services.
#clusvcadm ├в s ebrdo41d
#clusvcadm ├в s ecrde0q

As we have the outage window so i want to shutdown the cluster
# service rgmanager stop
# service clvmd stop
# service cman stop
# service qdiskd stop

Then i will take the backup
of /etc/cluster/cluster.conf

As per your advice i will do the following changes to cluster.conf

#vi /etc/cluster/cluster.conf [ as I already take the backup ]

---- increment the cluster config_version
[root@tneld52n ~]# grep config_version /etc/cluster/cluster.conf

[root@tneld52n ~]#

So i will make it 169
----as you have a two-node cluster, set the cman two_node parameter to 1 (=true)
--set the cman expected_votes parameter to 1

[root@tneld52n ~]# grep two /etc/cluster/cluster.conf

[root@tneld52n ~]#

So i will make two_node="1"
cman expected_votes="1"
- remove the entire block (i.e. all lines from line "", inclusive)

so i will delete the following lines

[root@tneld52n ~]# grep -i quo /etc/cluster/cluster.conf


[root@tneld52n ~]#

#ccs_tool update /etc/cluster/cluster.conf

As I already stop the cluster service. So just reboot the box. & if I see that it is up & cluster is running then I will copy the /etc/cluster/cluster.conf to other node .


ani007
Super Advisor

Re: Quorum disk removal

Hi Matti,
please find the attached Cluster.conf file in the above mail. & in this mail i am attaching the clustat o/p.
Matti_Kurkela
Honored Contributor

Re: Quorum disk removal

If the cluster is down, then the "ccs_tool update" command won't work... but that's OK, as you are allowed to edit the actual /etc/cluster/cluster.conf file when the cluster is down.

After the edits, the cman line should look like this:



When you're removing the quorumd block from the configuration, you should remove a total of 3 lines:





i.e. the line with the keyword "heuristic" should be removed too.

Before rebooting the nodes, remember to run "chkconfig qdiskd off" on each node, so that the now-useless qdiskd will not be started again when the node starts up.

MK
MK
ani007
Super Advisor

Re: Quorum disk removal

I copy the cluster.conf file to my home dir(/home/XXX/qdisk) & did the prework. sothat i can just copy paste the file at the outage window.Suppose i just want to check the file which i have prepaired now.I mean i will get the outage after 2 days. but i want to test it now . will it create any problem in my live production machine if i fire
#ccs_tool update /home/XXX/qdisk/cluster.conf.

Could you lease verify the two files. [/home/xxxx/qdisk/cluster.conf will be the file which i will place at the outage window]


[root@tneld52n qdisk]# diff /etc/cluster/cluster.conf /home/xxxx/qdisk/cluster.conf
2c2
<
---
>
24c24
<
---
>
146,148d145
<
<
<

[root@tneld52n qdisk]#
Matti_Kurkela
Honored Contributor

Re: Quorum disk removal

According to your diff output, your new configuration seems OK to me.

If you run

#ccs_tool update /home/XXX/qdisk/cluster.conf

it will implement the change immediately - in a single, synchronized, cluster-wide transaction. The new file is copied to all the nodes and the appropriate daemons notified of the change. If one of the nodes is unable to perform the configuration change, the change is automatically rolled back in all nodes.

The "ccs_tool update" command is not a test - it tells the cluster to actually make the configuration change.

If the heartbeat communication between the nodes works OK at the time of the change, I don't see why the removal of the quorum disk would interfere with the cluster services in any way.

In other words:
If you run the "ccs_tool update" command now and it's successful, the actual configuration change will be done without an outage, and you can use the full time of the outage for verifying that the cluster is rebootable in its new configuration, and other planned activities (if any).

If the "ccs_tool update" fails and does what it is documented to do at failure, no configuration change should happen and your cluster should keep running with its old configuration.

If there is a catastrophic failure for some reason, your boss will be unhappy because you did not wait for the scheduled outage...

When I was on a RedHat "Clustering and Storage" course, one of the standard exercises was to add a quorum disk to a running cluster. It worked. As I recall, the removal of the quorum disk was a simple reversal of the addition procedure.

In fact, the RedHat instructor told us that the use of the ccs_tool (or other configuration utilities) was actually preferable over editing the /etc/cluster/cluster.conf files manually. But the "ccs_tool update" only works when the cluster is running. The ccs_tool has other options which can be used to e.g. remove a node from the cluster.

MK
MK
ani007
Super Advisor

Re: Quorum disk removal

Thank you so much. I have successfully did it. Server is up & running fine.