Could not log into web GUI while other node was shut down?

[loneboat]

I have a 4-node cluster with nodes A, B, C, and D. I recently had to take nodes A and B offline to install a new NIC in them.

I wanted to watch their status on the web GUI of node C while I waited for them to boot back up, so after I shut down nodes A and B, I pulled up node C's web GUI. I was unable to log into node C (the GUI kept telling me that I was using an invalid pw?), so I attempted the same thing on node D, with the same behavior.

Once nodes A and B rebooted however, I was able to log into their web GUI just fine.

I'm vaguely aware of the concept of "fencing" when one sets up HA node configurations, however I do not think that I have set up any HA options in my cluster.

Is there a reason why I was unable to log into the web GUI while 2 of my nodes were offline? Is this a deliberate way to keep me from logging in and making (potentially dangerous) changes while half of the cluster was offline? Would I (likely) have been able to log in with only ONE of the nodes offline?

Thanks.

 

[UdoB]

Hello,

when two of four nodes are unavailable the remaining nodes are no longer more than 50%. You lost quorum.

Log in via ssh and

• look at the current state: pvecm status
• tell the system to temporarily accept that: pvecm expected 2

Then you should be able to use the Webgui again.

Best regards

 

[itNGO]

4-Node Cluster means at all time 3 Nodes need to be online and ready to call it a majority vote.
When 2 Nodes are down, your cluster is in a "failure" state and unpredicted behavior is expected....

So for your Cluster, 1 Node at a time is the maximum you can shutdown and maintenance without any hazzle....

 

[loneboat]

Hello,

when two of four nodes are unavailable the remaining nodes are no longer more than 50%. You lost quorum.

Log in via ssh and

• look at the current state: pvecm status
• tell the system to temporarily accept that: pvecm expected 2

Then you should be able to use the Webgui again.

Best regards

4-Node Cluster means at all time 3 Nodes need to be online and ready to call it a majority vote.
When 2 Nodes are down, your cluster is in a "failure" state and unpredicted behavior is expected....

So for your Cluster, 1 Node at a time is the maximum you can shutdown and maintenance without any hazzle....

Thanks both, this answers my question perfectly.

 

[nemocx]

Hello,

when two of four nodes are unavailable the remaining nodes are no longer more than 50%. You lost quorum.

Log in via ssh and

• look at the current state: pvecm status
• tell the system to temporarily accept that: pvecm expected 2

Then you should be able to use the Webgui again.

Best regards

Thanks for this info. I was able to login by setting pvecm expected to 1