Setting up CTDB on OCFS2 and VMs ...

Michael Adam obnox at
Sat Jan 3 04:15:27 MST 2015


On 2015-01-03 at 09:37 +0000, Rowland Penny wrote:
> On 03/01/15 00:36, Michael Adam wrote:
> >On 2015-01-02 at 22:32 +0000, Rowland Penny wrote:
> >>On 02/01/15 22:07, Martin Schwenke wrote:
> >>>
> >>>To make the recovery lock work properly your
> >>>cluster filesystem needs lock coherency.  We have already discussed this
> >>>several weeks ago:
> >>>
> >>>
> >>>
> >>>Nothing has changed.
> >>Yes, nothing has changed, the ping_pong test works just like the wiki page
> >>says it should,
> >So what does really happen? The following?
> >- ping_pong seems to work, i.e. you  run (e.g.)
> >   "ping_pong file 3" on two nodes and you see
> >   lock rates printed by both processes?
> >- "ping_ping -rw" seem sto work in that it
> >   prints correct data increment values
> Hi Michael, It has been some time since I last ran the ping_pong test, but
> when I did run it, I followed the wiki page to the letter and got the
> results that the page said I should.

Right, we had already come to the conclusion that
this is not sufficient.

> I have now run the test again, but in the way you suggested and you are
> correct, it isn't working, both nodes are printing lock rates.

Alright, that finally proves what we have been hinting
towards: Your OCFS2 setup is not correct yet.
Inter-node locking is missing.
Will update the wiki.

> I also see that Gunter Kukkukk has found the same webpage that
> I found that refers to NFS.

The mere fact that one can find a web page referring to
NFS does not make the hint more correct (in your use case).
I think we have explained at length why this does not
fix the flaws of the OCFS setup but just covers them up
and leaving you prone to run into more severe errors later on.

So where are we? We now kow for sure that you need to
do some changes to the OCFS2 setup. And other posts in
this thread (by Richard, Günter and Ralph) do have hints
how to get fcntl byte range locks working in OCFS.
(Especially see Günter's most recent posts.)

> Lets see how he goes on, or should he give up as well Martin?

Nobody should give up!

We have just reached a point where my earlier statements
is confirmed that ctdb is really easy to set up, and the
more difficult part is usually to correctly set up the
cluster file system (Note that "correctly" means correctly
for the needs of ctdb, you have certainly created a valid
OCFS2 setup, it is just not such that ctdb can work with it.)

All that being said, one final hint, in case you would give it
another try: You would probably make better progress if you
tried to listen more to what we say on this list instead of
trying to prove us or the ctdb software wrong by pointing to
resources on the web. We are really trying to help! :-)

Cheers - Michael
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: not available
URL: <>

More information about the samba-technical mailing list