[Ocfs2-users] re: question on adding a node to RAC cluster and o2cb

[Sunil Mushran]

Check the contents of /etc/ocfs2/cluster.conf. Are you sure there is no 
node number 2 in it.
Else email me a copy of the config file to see if I can duplicate it in 
house.

BTW, the directions for adding nodes (online & offline) have been added 
to the FAQ.
http://oss.oracle.com/projects/ocfs2/dist/documentation/ocfs2_faq.html#CONFIGURE

Peter Santos wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Sunil,
>   per your instructions I added our 3rd node to the cluster via o2cb_ctl.
>   The first time it didn't work because it said the number already existed.
>
>   dbo1:~ # o2cb_ctl -C -i -n dbo3 -t node -a number=2 -a ip_address=192.168.6.9 -a ip_port=7777 -a cluster=ocfs2
> 	o2cb_ctl: Node number "2" already exists
>
>   Not really sure why it said node number 2 existed. I checked here.
> 	dbo1:~ # cat /config/cluster/ocfs2/node/dbo1/num
> 	0
> 	dbo1:~ # cat /config/cluster/ocfs2/node/dbo2/num
> 	1
>
>   I even checked with "o2cb_ctl -I -type node...
>
>   Then I re-ran my o2cb_ctl command with "-a number=3" and it worked.  Now I have node 0,1, and 3.  I don't suspect
>   this is a big deal, but I'm not sure why it didn't like node number 2 .. Any ideas ?
>
>   BTW, this time I was able to mount the ocfs2 filesystem without a problem... this was the whole reason for this
>   thread.. previously I had tried to add a node with ocfs2console and propagation .. but it didn't work until I shutdown
>   o2cb.
>
>
> - -peter
>
>
>
> Sunil Mushran wrote:
>   
>> The real error was the one you got when you were not able
>> to add the new node in node1. It is an ocfs2console problem.
>> That it did not work when you added the node in node2 and
>> propagated, is explainable.
>>
>> When you get the third node, do the following:
>>
>> 1. On the existing two nodes, add the new node by hand by
>> executing this (on both).
>> # o2cb_ctl -C -i -n NODENAME -t node -a number=NODENUM -a ip_address=IPADDR
>>          -a ip_port=7777 -a cluster=CLUSTERNAME
>>
>> 2. By doing so, you are not only adding the node in /etc/ocfs2/cluster.conf
>> but also activating it (/config/cluster/CLUSTERNAME/node).
>>
>> 3. Either "Propagate" or hand copy the cluster.conf to the new node.
>>
>> 4. Start the cluster on the new node and then mount.
>>
>> Peter Santos wrote:
>>
>> I don't know what the entries looked  like in
>> /config/cluster/<clustername>/node/
>> when we tried this.
>>
>>
>> Now it does show both nodes... but we have since restarted the entire
>> cluster in order to get this to work.
>> We are waiting to get another new machine to try it again.
>>
>> What I do remember is that initially we started up the ocfs2console
>> from node1 and
>> clicked "add" to add a 2nd node and the tool complained ( I can't
>> remember the exact error message now).
>>
>> Then we tried to run ocfs2console from the new/2nd node and added both
>> node1 and node2 to the configuration
>> Then we clicked propagate .. this worked without any error messages,
>> but we were not able to mount the ocfs2
>> filesystem on node2 until we restarted the cluster on node1.
>> (transport endpoint errors..)
>>
>> We will definitely try again on a 3rd node, I'm just not clear on what
>> the sequence of events
>> should be.
>>
>> thanks
>> peter
>>
>>
>>
>> Sunil Mushran wrote:
>>  
>>
>>     
>>>>> When you added the new node using ocfs2console, did it show up in:
>>>>> # ls /config/cluster/<clustername>/node/
>>>>>
>>>>> I am assuming that it was added in /etc/ocfs2/cluster.conf.
>>>>>
>>>>> Yes, the docs does not cover this as of now. I will update the
>>>>> FAQ/user's guide
>>>>> with the info.
>>>>>
>>>>> Peter Santos wrote:
>>>>>
>>>>> Folks,
>>>>>     I'm trying to find information about how to dynamically add
>>>>>     a 2nd node to a 1 node RAC cluster.  I'm posting this only after
>>>>> not getting the
>>>>>     details from my oracle tar via metalink.
>>>>>
>>>>>     My installation is Suse Enterprise 9 x86_64 (kernel 267).
>>>>>
>>>>>     Installing the single node was not a problem, what is not clear is
>>>>> how to prepare
>>>>>     the cluster.conf file and the ocr stuff to add a 2nd or additional
>>>>> node. Obviously the
>>>>>     2nd node has to have all the ip configurations in place and ssh
>>>>> has to be working, but
>>>>>     at some point, the /etc/ocfs2/cluster.conf file has to be modified
>>>>> and propagated and the
>>>>>     ocfs2 mount point has to be mounted on the additional nodes ..this
>>>>> is where we had problems.
>>>>>
>>>>>     Here is what we did.
>>>>>         1. setup the 2nd node with all the proper network
>>>>> configuration, and ssh equivalence.
>>>>>         2. we "added" a 2nd node to cluster.conf via ocfs2console and
>>>>> propagated that to the new node.
>>>>>         3. We tried to mount the ocfs2 mount point, but could not ..
>>>>> it said something like
>>>>>         "transpoint end point not found ...".
>>>>>
>>>>>         4. We then restarted the cluster on node1 and were able to
>>>>> mount the ocfs2 mount point and go
>>>>>            on to add the 2nd node.
>>>>>
>>>>>     We are trying to identify the sequence of actions/procedures to
>>>>> add a 2nd node at the o2cb/ocfs2 level.
>>>>>
>>>>>     Oracle support didn't have this level of detail, so I'm hoping
>>>>> someone knows how to do this without
>>>>>     shutting down the cluster on node1
>>>>>
>>>>>     thanks
>>>>>
>>>>> -peter
>>>>>
>>>>>     
>>>>>           
>> _______________________________________________
>> Ocfs2-users mailing list
>> Ocfs2-users at oss.oracle.com
>> http://oss.oracle.com/mailman/listinfo/ocfs2-users
>>
>>     
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.1 (GNU/Linux)
> Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
>
> iD8DBQFE6ha/oyy5QBCjoT0RArV2AJ9QuavpCzJaj7EhT/JkCsK01zgVzwCfQDrN
> nXeJVtvfApOyhoqXT6xHJ6g=
> =vuGQ
> -----END PGP SIGNATURE-----
>