Hi Adam,
The procedure you describe looks good.
Make sure the recovery log resync operation is correct.
There are a couple of resources describing how to restore a failed
controller:
-
http://sequoia.continuent.org/doc/infocenter/index.jsp?topic=/org.continuent.sequoia.doc/html/Recover_from_a_controller_node_failure.html
-
http://sharengo.org/infra/haplatform/docs/Sequoia/en/html_single/index.html
Let us know if you find an unsupported case.
Thanks for your feedback,
Emmanuel
It might be easier just to send you the config setup I have, but I
would rather not do that to this list.
I can see how it works if you have one controller with X backends. But
in the case where you have 2 controllers with 2 backends it did not
seem to work.
Here is what I did:
- Start both controllers
- Create a dump on C1 and copy it to C2
- enable backends
- Perform some db changes (At this point everything is in sync)
- kill c2 to emulate controller going down
- transfer and restore dump and recovery log
- write to c1
- enable c2 backend
I did not see the change get copied over. I will play more with this
as it is possible that something I did messed things up but your
saying that the above sequence of events should work fine and that the
extra change that did not get copied over in the recovery log restore
should be copied over once the backend is enabled.
------------------------------------------------------------------------
> Date: Thu, 10 Jul 2008 01:15:54 +0200
> From: [EMAIL PROTECTED]
> To: [email protected]
> Subject: Re: [Sequoia] DB failover does not seem to be working
>
> Hi Adam,
>
> I am not sure what your configuration really is and what exact commands
> you are issuing but the writes happening during a log transfer or
resync
> are appended to the recovery log and replayed when the backend is
> enabled. So when the backend is enabled it should be fully synchronized
> unless your recovery logs are corrupted (which should never happen).
> If you find a use case that is not working, please don't hesitate to
> file a JIRA entry.
>
> Thanks for your feedback,
> Emmanuel
>
>
> > Ok, from what I can tell you databases go out of sync..... thats not
> > good considering the usage our dbs are likely to get...... Is there a
> > way to solve this that I have not spotted that does not involve
taking
> > both controllers offline inorder to resync them.
> >
> >
> >
> >
------------------------------------------------------------------------
> > From: [EMAIL PROTECTED]
> > To: [email protected]
> > Subject: RE: [Sequoia] DB failover does not seem to be working
> > Date: Wed, 9 Jul 2008 15:52:30 -0400
> >
> > Well I seem to have most of it working finally after some tweaks.
> >
> > I do have one question I want to confirm however:
> >
> > I have 1 server running and 1 down and I bring up the second server,
> > transfer a dump over and then start the second backend. What happens
> > if in between the time I took the dump and the time backend 2 comes
> > online a write was made to the 1st database. How do we ensure that
the
> > second one is up to date (the two databases in this case are replicas
> > of each other.).
> >
> > I am just interested in what is meant to happen. I have managed to
> > successfully restore the second backend and it gets writes made after
> > it is running.
> >
> >
> >
> >
> >
------------------------------------------------------------------------
> >
> > From: [EMAIL PROTECTED]
> > To: [email protected]
> > Date: Mon, 7 Jul 2008 13:41:56 -0700
> > Subject: Re: [Sequoia] DB failover does not seem to be working
> >
> > Hi Adam,
> >
> > You need to restore the backup on the second controller. The general
> > flow is as follows:
> >
> > 1.) Initialize first backend.
> > 2.) Dump backend contents. This creates a dump with an associated
> > checkpoint. You can enable the backend at this time.
> > 3.) Transfer log and dump to second controller. This gets the
> > controllers syncrhonized.
> > 4.) Load the dump on the second backend.
> > 5.) Enable the second backend.
> >
> > At this point your cluster is up and running.
> >
> > Cheers, Robert
> >
> > On 7/7/08 12:27 PM, "Adam Purkiss" <[EMAIL PROTECTED]> wrote:
> >
> >
> > That was the reason it did not start up.
> >
> > I will see if this fixes things
> >
> > Can I confirm that if I use this then when I restore backend DB1
> > on controller 1 do I need to restore the same backup on controller
> > 2 or do I just enable it.
> >
> >
> >
> >
> >
> >
> > From: [EMAIL PROTECTED]
> > _To: [EMAIL PROTECTED]
> > _Date: Mon, 7 Jul 2008 11:54:09 -0700
> > Subject: Re: [Sequoia] DB failover does not seem to be working
> >
> > Hi Adam,
> >
> > Can you check your hedera.properties file and look at the JGroups
> > config file name? If it is sequencer.xml, you have just hit a
> > configuration problem that is logged as sequoia-1102, which is
> > fixed in the codeline. The problem is that the jgroups-all.jar
> > file also contains a sequencer.xml file with a bad IP address. It
> > picks this up before the Sequoia version.
> >
> > To correct the problem, rename config/sequencer.xml to
> > config/sequoia_sequencer.xml and adjust the name accordingly in
> > the Hedera properties file referenced by your virtual database.
> >
> > Thanks, Robert
--
Emmanuel Cecchet
FTO @ Frog Thinker
Open Source Development & Consulting
--
Web: http://www.frogthinker.org
email: [EMAIL PROTECTED]
Skype: emmanuel_cecchet
_______________________________________________
Sequoia mailing list
[email protected]
https://forge.continuent.org/mailman/listinfo/sequoia
