What to do when the Cluster Master failed?

[klabauter]

What to do when the Cluster Master failed?

I saw you could make one node from a slave to a master, but what ist happening when the old master comes back zu service?

It is possible to put the master in in automatic failover environment ?

 

[dietmar]

What to do when the Cluster Master failed?

I saw you could make one node from a slave to a master, but what ist happening when the old master comes back zu service?

That is a bad idea - remove the old master before you enable a new master.

It is possible to put the master in in automatic failover environment ?

will be in V2.0

 

[klabauter]

How to remove the old master?
Do I have to execute following command on every slave ?

pveca -d MASTERID

then executing

pveca -m

on the new master?

Would be nice to see this issue explained in the wiki at the cluster page.

 

[hk@]

hi
I agree - this is a bit - well - undocumented.

I tried to promote another host to master, now this host thinks it is the master, the real master has been removed from it's node-table and the other five hosts think it (the new master) is still a node and belive the old master is their master...

please advise - regards
hk

 

[dietmar]

I tried to promote another host to master, now this host thinks it is the master, the real master has been removed from it's node-table and the other five hosts think it (the new master) is still a node and belive the old master is their master...

You need to tell the nodes that there is a new master. Execute this on each node:

# pveca -s -h <ip_of_new_master>

 

[whinpo]

hi
I agree - this is a bit - well - undocumented.

The wiki is opened to users willing to help to write documentations...

 

[hk@]

agreed - but how could I write a documentation if I wouldn't know the proper way first hand...