DRBD replication for disaster Recovery

[niru.feeds]

Hello All,

I have tested DRBD with HA (same network clustering), but I wanted to know how can I use DRBD (2 nodes) in different locations (local & remote) for disaster recovery.

Scenario:

we have a small data center with 20 servers (including blades and VMS) all of these servers use iSCSI SAN for storing data. I have a identical setup in my DR site. some of the servers are used on other networks (VPNs, Branch etc.)

We also have a network redirection in case of failure. That means if the primary (production) network goes down, all the traffic will be routed to our DR site. as well we have few client systems (200 in primary location and 10 in DR site) which use these servers. At the moment we are taking tape backups of each logical disks in primary SAN and putting them on the DR site SAN. thats time consuming and with so many logical volumes too many tapes are being used.

As I mentioned in the primary site we have tested DRBD with HA and it works, all the iSCSI targets are available to the primary site servers through the DRBD failover mechanism.

What I want is take one of the DRBD servers and keep it in DR site (connected via wireless / OF) and do the replication.

I found that if I stop the heartbeat, the replication between the nodes stops and with remote networks, how do I do the heartbeat. our remote site is different subnet. we configured IP redirection, even DNS change if primary goes down, all in all, we have network switchover in case "primary" fails.

I need help in doing the DRBD without HA configuration (which needs either cross over or same subnet) at the same time the replication should be both ways with latest data replicated on both DRBD servers.

Please let me know how can I go for this.

[Adam Gandelman-2]

Hi-

If you've got a local HA cluster working and want to move one node to
the DR site, make the two nodes talk to each other over the WAN and
update drbd.conf (IPs, ports, protocol).  Take HA out of the equation
and manage DRBD manually (add to existing recovery procedure?) because
trusting a CRM with cross-site failover is not the best idea. See how it
works.

Or you can  add a third node to your existing setup. Stack it on top of
your existing DRBD resource and achieve both HA + geographically diverse
redundancy:
http://www.drbd.org/users-guide/s-pacemaker-stacked-resources.html

Performance over your WAN may be sub-par, consider DRBD Proxy.

--
: Adam Gandelman
: LINBIT - Your Way to High Availability
: 8152 SW Hall Blvd., Suite #209 : Beaverton, OR 97008
: http://www.linbit.com




niru.feeds wrote:

> Hello All,
>
> I have tested DRBD with HA (same network clustering), but I wanted to know
> how can I use DRBD (2 nodes) in different locations (local & remote) for
> disaster recovery.
>
> Scenario:
>
> we have a small data center with 20 servers (including blades and VMS) all
> of these servers use iSCSI SAN for storing data. I have a identical setup in
> my DR site. some of the servers are used on other networks (VPNs, Branch
> etc.)
>
> We also have a network redirection in case of failure. That means if the
> primary (production) network goes down, all the traffic will be routed to
> our DR site. as well we have few client systems (200 in primary location and
> 10 in DR site) which use these servers. At the moment we are taking tape
> backups of each logical disks in primary SAN and putting them on the DR site
> SAN. thats time consuming and with so many logical volumes too many tapes
> are being used.
>
> As I mentioned in the primary site we have tested DRBD with HA and it works,
> all the iSCSI targets are available to the primary site servers through the
> DRBD failover mechanism.
>
> What I want is take one of the DRBD servers and keep it in DR site
> (connected via wireless / OF) and do the replication.
>
> I found that if I stop the heartbeat, the replication between the nodes
> stops and with remote networks, how do I do the heartbeat. our remote site
> is different subnet. we configured IP redirection, even DNS change if
> primary goes down, all in all, we have network switchover in case "primary"
> fails.
>
> I need help in doing the DRBD without HA configuration (which needs either
> cross over or same subnet) at the same time the replication should be both
> ways with latest data replicated on both DRBD servers.
>
> Please let me know how can I go for this.
>  

--
Adam Gandelman - 503-573-1262 x203
LINBIT - Your Way to High Availability
8152 SW Hall Blvd., Suite #209 : Beaverton, OR 97008

http://www.linbit.com

_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

[niru.feeds]

what do u mean my "manage DRBD manually" ?

and what is CRM ?

kindly put more light of these.


I've tried removing HA (stopping HA service) and putting one of the systems as primary, I dont get the iscsi targets.

regards
Niranjan

Adam Gandelman-2 wrote:

Hi-

If you've got a local HA cluster working and want to move one node to
the DR site, make the two nodes talk to each other over the WAN and
update drbd.conf (IPs, ports, protocol).  Take HA out of the equation
and manage DRBD manually (add to existing recovery procedure?) because
trusting a CRM with cross-site failover is not the best idea. See how it
works.

Or you can  add a third node to your existing setup. Stack it on top of
your existing DRBD resource and achieve both HA + geographically diverse
redundancy:
http://www.drbd.org/users-guide/s-pacemaker-stacked-resources.html

Performance over your WAN may be sub-par, consider DRBD Proxy.

--
: Adam Gandelman
: LINBIT - Your Way to High Availability
: 8152 SW Hall Blvd., Suite #209 : Beaverton, OR 97008
: http://www.linbit.com




niru.feeds wrote:
> Hello All,
>
> I have tested DRBD with HA (same network clustering), but I wanted to know
> how can I use DRBD (2 nodes) in different locations (local & remote) for
> disaster recovery.
>
> Scenario:
>
> we have a small data center with 20 servers (including blades and VMS) all
> of these servers use iSCSI SAN for storing data. I have a identical setup in
> my DR site. some of the servers are used on other networks (VPNs, Branch
> etc.)
>
> We also have a network redirection in case of failure. That means if the
> primary (production) network goes down, all the traffic will be routed to
> our DR site. as well we have few client systems (200 in primary location and
> 10 in DR site) which use these servers. At the moment we are taking tape
> backups of each logical disks in primary SAN and putting them on the DR site
> SAN. thats time consuming and with so many logical volumes too many tapes
> are being used.
>
> As I mentioned in the primary site we have tested DRBD with HA and it works,
> all the iSCSI targets are available to the primary site servers through the
> DRBD failover mechanism.
>
> What I want is take one of the DRBD servers and keep it in DR site
> (connected via wireless / OF) and do the replication.
>
> I found that if I stop the heartbeat, the replication between the nodes
> stops and with remote networks, how do I do the heartbeat. our remote site
> is different subnet. we configured IP redirection, even DNS change if
> primary goes down, all in all, we have network switchover in case "primary"
> fails.
>
> I need help in doing the DRBD without HA configuration (which needs either
> cross over or same subnet) at the same time the replication should be both
> ways with latest data replicated on both DRBD servers.
>
> Please let me know how can I go for this.
>  


--
Adam Gandelman - 503-573-1262 x203
LINBIT - Your Way to High Availability
8152 SW Hall Blvd., Suite #209 : Beaverton, OR 97008

http://www.linbit.com

_______________________________________________
drbd-user mailing list
[e-mail]
http://lists.linbit.com/mailman/listinfo/drbd-user