Cannot delete ghost node

[javierib]

Hello,
We have 3 servers in cluster. We had a problem adding a 4th server (it stayed in the "Waiting for quorum" status)...after rebooting all servers and disconnecting the last server we could gain access to the cluster again.
The problem now is that we have a ghost node with the "Estranged" status...what does it mean?, how can we delete that node? (it does not exist in any cluster.conf)
All nodes have the same and latest version (2.3/7946f1f1).

This is the summary screen: (the node that never joined the cluster was named proxmox2-5)



Any ideas will be appreciated...

Thanks in advance

 

[dietmar]

# rm -rf /etc/pve/nodes/Node4

 

[javierib]

The /etc/pve/nodes directory, in the 3 nodes of the cluster, contains the same:

root@proxmox2-0:/etc/pve/nodes# ls -lh
total 0
drwxr-x--- 2 root www-data 0 oct 9 2012 proxmox2-0
drwxr-x--- 2 root www-data 0 oct 9 2012 proxmox2-1
drwxr-x--- 2 root www-data 0 oct 9 2012 proxmox2-3
root@proxmox2-0:/etc/pve/nodes#

Also, clustat output gives:

root@proxmox2-0:/etc/pve/nodes# clustat
Cluster Status for clusterherrera @ Fri May 10 09:43:13 2013
Member Status: Quorate

Member Name ID Status
------ ---- ---- ------
proxmox2-0 1 Online, Local
proxmox2-1 2 Online
proxmox2-3 3 Online
Node4 4 Online, Estranged


We don't know why Node4 says it's online...the node that failed joining the cluster was reinstalled with another IP...we don't want to try to join again until the "Estranged" node problem resolves.

Thank you very much!

 

[javierib]

Very sorry for the duplicate posts....wasn't reading the message after posting

 

[dietmar]

the node that failed joining the cluster was reinstalled with another IP

You really whipped the whole disk (or how did you reinstall)?

 

[javierib]

You really whipped the whole disk (or how did you reinstall)?

I really whipped...fresh debian install, repartitioned and reinstalled with new IP.

 

[dietmar]

I really whipped...fresh debian install, repartitioned and reinstalled with new IP.

What is the output of

# pvecm nodes

 

[javierib]

dietmar,
thank you very much for your help....we've found the culprit...it was a previous node that was removed (proxmox2-2) and wasn't turned off