Bug 1395639

It turns out we cover this very issue in the RHEL 6 Pacemaker manual, in section A.2:

https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/7/html/High_Availability_Add-On_Reference/s1-pacemaker65-70-HAAR.html#s2-pmakercreate65-70-HAAR

This section of the RHEL 7 manual was written during 6.5 and not updated. I don't know if there is still a need for this particular section at this point -- are there new RHEL 7 users who are coming from 6.5? -- but on the other hand it seems simple enough to note this difference between 6.5 and all other releases that support Pacemaker and I'm reluctant to drop it entirely.

I would also say that it's not a problem to mention RHEL 6 in this document -- in the appendix in particular -- because the audience is still RHEL 7 users, but RHEL 7 users who are new to RHEL 7 but have been working with Pacemaker on RHEL 6 might need this info.  

Also the other section of the Appendix here is about installation on RHEL 6.5 vs RHEL 7, and while what it says about 6.5 is true for RHEL 6.6 and later as well it still talks only about 6.5, so I'll just note "6.5 and later" now.

Thanks for pointing this out.

According the Chris, we no longer support RHEL 6.5 anyway, so I think I can leave in the subsection on installation differences between RHEL 6 and RHEL 7 but I can remove the subsection on differences in cluster setup between RHEL 6.5 and all later releases.

Section A.2 of the RHEL 7 Pacemaker manual now has only one subsection -- on the differences in installation between RHEL 6 and RHEL 7. I have eliminated the subsection on the differences in cluster creation between RHEL 6.5 and later releases, as per Comment 2.

Chris:  For sanity's sake, could you look at Section A.2 here and let me know if this is ok?  This doesn't have to wait until 7.4, I can build this and push it to the Portal as soon as you approve this.

http://jenkinscat.gsslab.pnq.redhat.com:8080/job/doc-Red_Hat_Enterprise_Linux-7-High_Availability_Add-On_Reference%20%28html-single%29/lastSuccessfulBuild/artifact/tmp/en-US/html-single/index.html#s1-pacemaker65-70-HAAR

Steven

To match the RHEL 6 pacemaker installation more closely we can just highlight the only differences between RHEL 6 & RHEL 7 pacemaker setup.

Currently you have this:
[root@rhel6]# yum install pacemaker cman
[root@rhel6]# yum install pcs
[root@rhel6]# chkconfig corosync off

But from the RHEL6 pacemaker doc, we want this:
[root@newnode ~]# yum install -y pacemaker cman pcs
[root@newnode ~]# chkconfig corosync off
[root@newnode ~]# passwd hacluster
[root@newnode ~]# service pcsd start
[root@newnode ~]# chkconfig pcsd on

Then on both RHEL 6 & RHEL 7 you do the pcs cluster auth.


(This is the doc that I was referring to)
https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/6/html-single/Configuring_the_Red_Hat_High_Availability_Add-On_with_Pacemaker/index.html#s2-nodeadd-HAAR

(Making this needinfo on the BZ to keep the record here)

Chris:

It looks as though I need to make things consistent in a couple of places.  The example configuration we provide in the RHEL 6 Pacemaker manual uses this:

(https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/6/html-single/Configuring_the_Red_Hat_High_Availability_Add-On_with_Pacemaker/index.html#s1-installstartcluster-HAAA)

yum install -y pcs pacemaker cman fence-agents

So I don't know if at this point we need to install fence-agents explicitly

And this:

# chkconfig corosync off
# chkconfig cman off

That is -- in addition to chkconfig corosync off you also need to chkconfig cman off.  So I'm not sure about that either any longer. I know I tested all this a long time ago.

Chris:

And also, coincidentally, there's a new open BZ about the documentation for this very procedure in RHEL 6 that I was planning to contact you about:

https://bugzilla.redhat.com/show_bug.cgi?id=1402152

That suggests we specifically enable pacemaker as well as pcsd, even though you can do that with pcs cluster enable.

So that's a slightly different issue, except that it emphasizes that over time and releases the various places we document these setup commands have grown inconsistent and I should look at them all again as a piece.

Steven

It isn't necessary to specify 'fence-agents' because cman already requires it.  So it is save to leave it off that line (but doesn't hurt anything if it does exist.

By default, cman & corosync are disabled by default (so the chkconfig corosync/cman off does nothing).  However, if people were upgrading from rgmanager clusters, they would be turned on, so we just have that line in there to disable them, just in case they had already turned them on.

Instead of running 'chkconfig pacemaker on' we have the users run 'pcs cluster enable --all'.  That runs the exact same command, but does it on all the nodes.

So I think the only place we need to make a change is to remove fence-agents from the RHEL 6 yum install & to make sure we specify 'chkconfig cman off' whenever we're using 'chkconfig corosync off'


However, as Sam talks about in bz#1402152 it may make sense to let users know how to check if pacemaker is enabled (which would start the cluster on reboot).

To do that you can do two different things:

Run 'pcs status' if the cluster is running.  At the bottom of the output you'll see 'Daemon Status'.  You should see pacemaker is 'enabled' (all other services should be disabled).

If the cluster is not running you can run this command:
[root@host-600 lib]# chkconfig --list pacemaker
pacemaker      	0:off	1:off	2:on	3:on	4:on	5:on	6:off

If 2,3,4 & 5 are on, the cluster is enabled.

I have updated section A.2.2 of the RHEL 6 Pacemaker manual with the information in Comments 4 and 7.

The issue of documenting whether pacemaker is enabled is a RHEL 6 issue (this is an appendix in RHEL 7 just showing the installation differences so people used to RHEL 6 are not confused), so I'll put that information in the RHEL 6 document as part of bz#1402152.

In the meantime, here is the updated section that this BZ refers to:

http://jenkinscat.gsslab.pnq.redhat.com:8080/job/doc-Red_Hat_Enterprise_Linux-7-High_Availability_Add-On_Reference%20%28html-single%29/lastSuccessfulBuild/artifact/tmp/en-US/html-single/index.html#s1-pacemaker65-70-HAAR

If you could check off on that I can move this bug forward...

Looks good to me.