Cluster Membership - Removing / Adding Server Instances

Users
1

Because of the issues I’ve had with drop all, I tried to just delete the various database files (p & w directories) and restart the servers. I got several errors when the servers came back up and tried to rejoin (eventually they failed to join).

My procedure was:

  1. Stop the docker container for each server.
  2. Remove the p and w directories for each server.
  3. Start the docker container for each server.

Since that didn’t work and this is a development test setup, I tried completely recreating the server containers but rain into an issue when the servers tried to join the cluster. They couldn’t join because zero already had a member with the same name. I had to completely recreate the cluster.

I realize I’m hacking my way through this the hard way so the fact that things aren’t happy isn’t a giant surprise.

How do you remove a server from the cluster completely?

If you have a hard failure and need to recreate a server instance, assuming you have the data files to restore, how do you allow a new container image for a server to join the cluster under the same name/port combination?

2

See the /removeNode API at https://docs.dgraph.io/deploy/#more-about-dgraph-zero.

Since you are going the way of removing the directories and restarting, I’d suggest do the same for Zero as well i.e. remove the zw directory. In that case Zero won’t complain.

3

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.