Behavior on disk errors

[Tim Westbrook]

We are trying to understand the behavior of drbd when disk errors occur in a few cases.

1) When a separate metadata disk is created, what happens when i/o errors happen on that disk? 

Do the node go into Diskless? 


2)  When a node is a secondary, what happens when drbd gets i/o errors? 

I understand primary will continue to stream data to the remotes when errors are detected (by default) but what will happen on a secondary if errors are detected, Does it go into a disconnected state?

Cheers,
Tim