Best way to replace a node?

DRBD
,
1

Hello,

I have a working NFS - HA - DRBDB v9 with reactor (primary / secondary and one arbiter diskless).
I want to change the secondary by another server (exactly same hardware except the location in datacenter) without losing data and access to NFS for client (no downtime).

My plan :

  • evict the secondary node B with IP 192.168.1.X and changes IP adress to 192.168.1.Y
  • install new server with IP 192.168.1.X
  • initialize DRBD ressource on the new server it’will connect to primary and resync right ?

Or it’s better to change configuration and use another IP ?

I’m worried about reloading the reactor or drbd configuration and crashing the NFS service.

Or there is a better way to acheieve this ?

Thanks

Yathus

2

Your plan seems reasonable to me. Copy all the configuration files over to the new system, give the new system to same IP address and you should be pretty good to go.

A few things to keep in mind would be to make sure to keep the version numbers the same. It’s very easy for a new deployment to install some newer version by default. Additionally, I would probably bring up DRBD manually alone without Reactor first and let the initial sync complete before starting Reactor and fully introducing the new node into the cluster. This shouldn’t be needed, but I tend to be overly cautious.

1 Like
3

In case it might also be helpful, this LINBIT blog post has some general instructions for how to perform system maintenance tasks while minimizing downtime, in both Pacemaker and DRBD Reactor clusters.

1 Like