On Feb 18, 2008 7:33 PM, Eddie C <[EMAIL PROTECTED]> wrote:
> The reason I began researching this type of replication is that we had
> several applications that could not be easily clustered because they
> had some persistant data such as logs and state information that was
> stored locally on disk. So it seemed like the best way was to create
> drbd disks, I went with the active/active disk configuration in DRBD8,
> and putting OCFS2 on top of it.

If you merely need a persistent data store for each application, why
not just make a DRBD resource with ext3/xfs on top for each
application, and let Heartbeat colocate the each DRBD resources and
application? This is basically the sort of setup we are running now.


> First note that as a rule of thumb I never build custom kernels. Maybe
> I am too new school, but I figure if the application does not work
> with the kernel I am running now it is very bleeding edge and probably
> not ready for mainstream deployment.
>
> I did find that OCFS2 and or DRBB 0.8 would not work with the stock
> FC5 kernel. A yum update to the newest kernel got the system to the
> point of running the software in memory without Dynamic Link issues.
> Firstly I had to handle the drbd partitioning. This is not a big deal
> but, its not 1,2,3. You can not really kickstart it well. have to
> configure OCFS2, have to open firewall ports, also add a crossover
> cable, put the system on two networks.
>
> But I did get it up and working in active/active mode. I should have
> ran UnixBench on it to see the raw performance but I did not.
>
> Also at the time the OCF could not handle active/active drbd
> configurations. I hacked together some LSB scripts for DRBD, ocfs2.
>
> All in all the setup time was vast. I felt that I could get better
> value out of one machine well administered and backed up. the major
> problem is that you had to be an HA expert, a DRBD expert, and an OCFS
> expert to get anything working on the systems.
>
> I think the active active ocfs2 would be great for a large SAN type
> disk, but it was just too much management for servers you would like
> to replicate.
>
>
> On Feb 18, 2008 11:39 AM, Atul Athavale <[EMAIL PROTECTED]> wrote:
> > Hello,
> >
> > You may find some starting point at
> >
> > http://wiki.linux-ha.org/AtulAthavale
> >
> > or at
> >
> > http://www.debian-administration.org/articles/578
> >
> > Regards,
> >
> > Atul Athavale.
> >
> > _______________________________________________
> > Linux-HA mailing list
> > [email protected]
> > http://lists.linux-ha.org/mailman/listinfo/linux-ha
> > See also: http://linux-ha.org/ReportingProblems
> >
> _______________________________________________
> Linux-HA mailing list
> [email protected]
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
>
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to