How to: cold standby server (no cluster)

[SiliconJunkie]

Originally Posted by leog21

We've successfully setup Zimbra on a DRBD / Heartbeat solution with automatic failover + one-to-one NAT. I'm currently working on documenting an implementation plan.

Best,
Leo

Anxiously waiting to read about your project. I need to do this very thing right now, and would appreciate relating to your experiences.

Cheers!

[leog21]

You need to know how to setup DRBD and HeartBeat, that I can not cover in this thread.

we have two boxes, Node1 and Node2, and a Virtual IP by heartbeat

install Zimbra on both disks

then DRBD the /opt partition, zimbra should be working on either node1 or node2 depending on which one is primary or secondary

Setup heartbeat to monitor the VIP and bring up zimbra in case of failover
i.e. /etc/ha.d/haresources/ on the first line type

node1 drbddisk Filesystem::/dev/drbd0::/opt::ext3 x.x.x.x(your VIP ip address) zimbra

This can also be done by hand, you don't need to have heartbeat.

/opt will switch over between the two nodes

As far as one-to-one Nat, point your public ip address to the vip address

Add. Notes:
- I recommend identical boxes for both nodes to make things easier
- Upgrading Zimbra is simple, just run the upgrade on the Primary node (active). I've performed an upgrade from 4.5.7 to 5.0.5 with no problems thanks to the amazing upgrade process Zimbra has.
- Shutdown/Restart

Shutdown:
1 stop ha on node2
2 stop drbd on node2  ->now shutdown node2
3 stop ha on node1
4 stop drbd on node1  ->now shutdown/restart node1
 
Restart:
-> turn on node1
5 start drbd on node1 (can't find the other, bypass, say "yes")
6 start ha on node1 (auto)
-> turn on node2
7 start drbd on node2 (auto)
8 start ha on node2 (auto)

YOU NEED TO STUDY DRBD or else you risk hosing your /opt partition.
Also look into the SPLIT BRAIN PROBLEMS that DRBD may encounter, thats when both nodes don't know who will take over the main partition. Please practice messing around with DRBD before you put a production system online running a DRBD parition.

I hope this helps. Sorry I can not be any more detailed at the moment.

[parixit]

What happened to the server to server sync that was supposed to be in V. 6.0?

[phoenix]

Originally Posted by parixit

What happened to the server to server sync that was supposed to be in V. 6.0?

I don't know what your question has to do with this thread but you'll find details in the bug report and the Product Portal for the schedule for bugs.

[parixit]

Unless i have mistaken (which probably is the case) but wouldn't that take care of having a cold stand by server?

[iway]

@phoenix: OK, but a more complex answer would be appreciated here.

[iway]

@parixit: Yes, it should make the cold standby obsolete by being even better...

[parixit]

@iway I agree. I was really looking forward to it being in the 6.0 release. Are they even planning on including it?

[iway]

@parixit: I know, this would have been one of the most-needed features! No idea when it will be included, though.
Zimbra don't give any information, asking them gets you no reply at all.

@Zimbra: Please provide a roadmap and an estimated timeframe for this critical feature!

[raj]

as per the product portal Zimbra Product Portal under
RFEs with Most Votes (15) ?

11 11423 P1 disaster recovery through server to server sync (beta) At Risk (6.5)

looks like this feature is pushed to 6.5 and i am not sure what At Risk (6.5) means..does it means it need more votes ??
or it means it likely to be killed


Raj