One problem with the below workflow.
7. Your going to need to copy this over manually otherwise it
will fail, I've fallen victim of this before. All cluster nodes need to start on
the current revision of the file before you update it. I think this is a chicken
and egg problem.
One of
this things I have configured on my clusters is that all clustered services
start on it's own runlevel, in my case I have cluster services running on
runlevel 3 but default boot to runlelvel 2. This allows a node to boot up
and get network before racing into the cluster (ideal for wanting to find out
why it got fenced and solving the problem).
Everything else will work as I've just done this myself
(except 5 nodes). Your downtime should be quite
minimal.
Regards,
Peter Tiggerdine
HPC & eResearch Specialist
High
Performance Computing Group
Information Technology Services
University of
Queensland
Phone: +61 7 3346 6634
Fax: +61 7 3346 6630
Email:
peter.tiggerdine@xxxxxxxxx
From: linux-cluster-bounces@xxxxxxxxxx [mailto:linux-cluster-bounces@xxxxxxxxxx] On Behalf Of Gianluca Cecchi
Sent: Tuesday, 3 November 2009 7:29 AM
To: David Teigland
Cc: linux-cluster@xxxxxxxxxx
Subject: Re: share experience migrating cluster suite from centos 5.3 to centos 5.4
On Mon, Nov 2, 2009 at 6:25 PM, David Teigland <teigland@xxxxxxxxxx> wrote:
The out-of-memory should be fixed in 5.4:
https://bugzilla.redhat.com/show_bug.cgi?id=508829
The fix for dlm_send spinning is not released yet:
https://bugzilla.redhat.com/show_bug.cgi?id=521093
Dave
Thank you so much for the feedback.
So I have to expect this freeze and possible downtime...... also if my real nmodes a safer method could be this one below for my two nodes + quorum disk cluster?
1) shutdown and restart in single user mode of the passive node
So now the cluster is composed of only one node in 5.3 without loss of service, at the moment
2) start network and update the passive node (as in steps of the first mail)
3) reboot in single user mode of the just updated node, and test correct funcionality (without cluster)
4) shutdown again of the just updated node
5) shutdown of the active node --- NOW we have downtime (planned)
6) startup of the updated node, now in 5.4 (and with 508829 bug corrected)
This node should form the cluster with 2 votes, itself and the quorum, correct?
7) IDEA: make a dummy update to the config on this new running node, only incrementing version number by one, so that after, when the other node comes up, it gets the config....
Does it make sense or no need/no problems for this when the second node will join?
8) power on in single user mode of the node still in 5.3
9) start network on it and update system as in steps 2)
10) reboot the just updated node and let it start in single user mode to test its functionality (without cluster enabled)
11) reboot again and let it normally join the cluster
Expected result: correct join of the cluster, correct?
12) Test a relocation of the service ----- NOW another little downtime, but to be sure that in case of need we get relocation without problems
I'm going to test this tomorrow (here half past ten pm now) after restore of initial situation with both in 5.3, so if there are any comments, they are welcome..
Thanks
Gianluca
-- Linux-cluster mailing list Linux-cluster@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/linux-cluster
