The cluster has not any resources defined.

[half_life]

CA being active and configured wouldn't happen to be a pre-req would it?

[sixstone]

No, it is a shared-secret scheme.

FYI, the key exchange is crypted .

[half_life]

I am still where I was last time I wrote.  I am unable to get the cluster to do anything but be split brained.  I dunno and the logs aren't giving me any clues.  Anyone have any thoughts before I put this back down and wait for more time to play with it?

[sixstone]

Hello,

If I were you, I'd select a node and do as follows: Leave the cluster, join the cluster again and save changes. This should do the trick and you will not lose any information (conf replicated).

Only if you are interested in getting to know what's wrong, check the corosync auth file (/etc/corosync/authkey) is equal in both nodes and they are reachable each other. Running corosync-quorumtool -l and such it may help to debug the issue.

Best,

[peterpugh]

Dunno if anybody else fancies having a look at HA.

http://sourceforge.net/projects/zentyal/files/Incoming/

I have 3.4 loaded and so far its been a pretty painless affair. OPenChange and SOGo work great but still to have confirmation if anybody else struggles.

Samba seems to work but today I will be adding a few OU's and users with group policies. Just to do some further testing its a requirement of Openchange and I don't really expect anything undue.

I have my main system going with 3.4 and have already added the proxy which seems fine but haven't tried the SSO and filter groups yet.

Only thing seems to be that the bundle part of the openvpn server is missing, prob just an oversight with the new distro.

I will say that I really don't like Half-Lifes title of bughunter and it is nothing to do with Half-Life as is he is prob the best on here at it.
I think anyone can be a bug hunter and you don't need to get techy, just posting where at what things went wrong is enough.
If you can include a screen shot or better the log then that just add's more.
Being a noob is extremely important as your experience will tell much.

If there are any talented log sniffers out there who have limited time maybe if a bug is posted you could just confirm and post logs and stuff so its concise for the devs.

Half-Life I said it before about "Bug-Hunter" honestly we have a singular bughunter Doh! Lol wish they would give you a proper title maybe beta co-coordinator or QA co-coordinator. If titles are a must?
I cringe every time I see that title Our singular bug hunter seems to set a tone to me, but hey it is me!

Back to HA I deliberately created a separate thread on HA because I wanted to document from start to finish.
Hopefully even though a bit long winded it can act as a quick start guide to start.

With a little co-ordination I am sure many heads would make light work. I stopped at a point because I really don't understand how it can work without a quorum disk.
I am not saying it can not work without a quorum disk but personally I just don't understand how?

Would be great when it comes to HA that maybe someone would slap there meatware on the table and explain a little more.

Coffee, bacon butty, visit to the bank as low on supplies and will have a look at the info Sixstone has provided.

My reading when I get back http://clusterlabs.org/doc/en-US/Pacemaker/1.1-plugin/html/Clusters_from_Scratch/ch05s03.html

[peterpugh]

Hello,
Only if you are interested in getting to know what's wrong, check the corosync auth file (/etc/corosync/authkey) is equal in both nodes and they are reachable each other. Running corosync-quorumtool -l and such it may help to debug the issue.
Best,

node1 authkey
€áÓ¹Á|†•àêÀ›3˜šfÀ!Nç)WçÀ;ðrK
%Õ¨²Í,0õ­1´W±Xy‹+‹g@TôÖÞE|Ѹ ×ôˆ\‡´6/3"ؒ=~¶…å+; ’áH™éþĸ–:à¸mtwæxF¼Õä–É64oޓƒ^®YQ

node2 authkey
€áÓ¹Á|†•àêÀ›3˜šfÀ!Nç)WçÀ;ðrK
%Õ¨²Í,0õ­1´W±Xy‹+‹g@TôÖÞE|Ѹ ×ôˆ\‡´6/3"ؒ=~¶…å+; ’áH™éþĸ–:à¸mtwæxF¼Õä–É64oޓƒ^®YQ

node 1 corosync-quorumtool -l
Membership information
----------------------
    Nodeid      Votes Name
         1          1 192.168.3.1 (local)

node2 corosync-quorumtool -l
Membership information
----------------------
    Nodeid      Votes Name
         2          1 192.168.3.4 (local)

zentyal@zent1:~$ sudo corosync-quorumtool -s
Quorum information
------------------
Date:             Sun Mar  9 18:30:38 2014
Quorum provider:  corosync_votequorum
Nodes:            1
Node ID:          1
Ring ID:          56
Quorate:          No

Votequorum information
----------------------
Expected votes:   2
Highest expected: 2
Total votes:      1
Quorum:           2 Activity blocked
Flags:           

Membership information
----------------------
    Nodeid      Votes Name
         1          1 192.168.3.1 (local)


zentyal@zent2:~$ sudo corosync-quorumtool -s
Quorum information
------------------
Date:             Sun Mar  9 18:29:44 2014
Quorum provider:  corosync_votequorum
Nodes:            1
Node ID:          2
Ring ID:          16
Quorate:          No

Votequorum information
----------------------
Expected votes:   2
Highest expected: 2
Total votes:      1
Quorum:           1 Activity blocked
Flags:            2Node WaitForAll

Membership information
----------------------
    Nodeid      Votes Name
         2          1 192.168.3.4 (local)

Still confused as each quorum is expecting 2 votes but only getting one.
This is what I don't get if you havent got a shared quorum disk.
I presume there is some form but local to each machine which means that each node thinks its alone?!

This time before I down or reboot a VM I am going to remember to disable HA.

[peterpugh]

Had a browse around and still no thoughts then I suddenly thought "Hold on the port in Zentyal is the replication port", "What is the HA corosync port?"

So enabled 5405 UDP on both servers.

zentyal@zent1:~$ sudo corosync-quorumtool -s
Quorum information
------------------
Date:             Sun Mar  9 21:39:31 2014
Quorum provider:  corosync_votequorum
Nodes:            2
Node ID:          1
Ring ID:          148
Quorate:          Yes

Votequorum information
----------------------
Expected votes:   2
Highest expected: 2
Total votes:      2
Quorum:           2 
Flags:            Quorate

Membership information
----------------------
    Nodeid      Votes Name
         1          1 192.168.3.1 (local)
         2          1 192.168.3.4


zentyal@zent2:~$ sudo corosync-quorumtool -s
Quorum information
------------------
Date:             Sun Mar  9 21:35:54 2014
Quorum provider:  corosync_votequorum
Nodes:            2
Node ID:          2
Ring ID:          148
Quorate:          Yes

Votequorum information
----------------------
Expected votes:   2
Highest expected: 2
Total votes:      2
Quorum:           1 
Flags:            2Node Quorate WaitForAll

Membership information
----------------------
    Nodeid      Votes Name
         1          1 192.168.3.1
         2          1 192.168.3.4 (local)

Still got the red box on save though?

Looking at the logs not sure?

Also only a small thing but the demote and promote actions would seem to be the wrong way round.

Have a look at the image

[peterpugh]

Now I am confused, now what? what do I do?

[half_life]

I had understood your direction and had already tried leaving and rejoining to no avail.

Checking the files:
md5sum says they are the same.
dfbc5288d0e0d8d0ea79b7dc965fe91c  authkey-campion
dfbc5288d0e0d8d0ea79b7dc965fe91c  authkey-starkey

This is one unit:

Membership information                                                                                                                                                                                                                                                         
----------------------                                                                                                                                                                                                                                                         
    Nodeid      Votes Name                                                                                                                                                                                                                                                     
         2          1 192.168.163.4 (local) 




The other unit is currently frozen in a /etc/iniit.d/zentyal restart at the HA point in the sequence.  I suppose I will do another tear down and clean rebuild using the latest daily build on my next day off.


 I am very interested in this.

[peterpugh]

I was exactly the same with the joining node. If it fails do not reboot or it seems to get stuck. Disable HA first (passive node anyway).

Using a VM I just cloned the first and changed the IP and host name. Deleted the dead server.

Adding a service HA with UDP 5405 (both nodes)

Then allowing on the firewall.

I was looking on the HA screen and kept thinking 443 a funny port for a default.
It suddenly dawned on me this is purely for replication shipping and is the webadmin port, doh! obviously.
That is why it says blocked because it is.

My question is OK, It works and I have a floating IP.

But showing my complete ignorance in conjunction with services I am totally unsure what to do next?
External ports have me totally baffled?

Adding another float auto updates across nodes.

OK Euston, guide us in!

ps I was browsing http://clusterlabs.org/wiki/Debian_Lenny_HowTo

ps ps Community QA (much better title)

The leaving and rejoining.

The floating IP sets something somewhere.
On the designated node whilst module enabled.
Remove all floats.
leave cluster and create new.
add float

node2
join

[peterpugh]

I sort of had to think about HA again apols for being slow.

I am going to let Half-Life continue with his thread here.

I am going to jump back to my Beta HA post in the below url. Still a few things that I am unsure of.
https://forum.zentyal.org/index.php/topic,20921.msg80385.html#msg80385

Apols from jumping in but it was just to utilise the info from sixstone

[sixstone]

Hi all,

The other unit is currently frozen in a /etc/iniit.d/zentyal restart at the HA point in the sequence.  I suppose I will do another tear down and clean rebuild using the latest daily build on my next day off.

I suppose you allow all connections in your private network between both nodes, it is not that case. You must let heartbeat packets from corosync pass in the firewall. The corosync server is listening on UDP/5405 as peterpugh has explained above.

I'd better to let the firewall open between cluster nodes to leave entropy away.

Best regards,