[DRBD-user] Strange failure drbdmanage

Roland Kammerer roland.kammerer at linbit.com
Tue Feb 21 09:19:56 CET 2017

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.

On Mon, Feb 20, 2017 at 03:54:20PM +0100, Frank Rust wrote:
> Then I did "drbdmanage uninit" on the new node because I wanted to redo the configuration.
> From this moment on I get following on the first node:
> drbdmanage nodes
> Waiting for server: ...............
> No resources defined

This rendered it "unusable" because:
- The second control node was gone (uninit), but that does not mean it
  isn't in the control volume (aka. cluster DB) anymore.
- The first node has the second one in its DB and wants to do leader
- A pure satellite (without a control volume) does not count during
  leader election, so you effectively have a 2 node cluster where one
  node is gone.
- The first node will not start to avoid splits on the control volume.
  In a two node cluster both nodes have to be available.

What you should have done was to remove the second node while the
control volume was still available. "remove-node". Then your setup would
have become a 1 node setup from a leader election point of view (+ the
satellite), and would have started.

The easy way out is to also "uninit" the first node and start from
scratch. If the control volume already contains useful information like
lots of already configured drbd resources, you can force a win of the
leader election ("reelect --force-win"), "remove-node -f" the second,
"restart -q" to cleanly start it as a single node cluster, "add-node"
the second one.

Regards, rck

More information about the drbd-user mailing list