Status: new
Owner: ----
Labels: Type-Enhancement OpSys-All Performance
New issue 381 by [email protected]: Drain that doesn't force
master_candidate=no
http://code.google.com/p/ganeti/issues/detail?id=381
What software version are you running? Please provide the output of "gnt-
cluster --version" and "gnt-cluster version".
# gnt-cluster version
Software version: 2.6.1
Internode protocol: 2060000
Configuration format: 2060000
OS api version: 20
Export interface: 0
# gnt-cluster --version
gnt-cluster (ganeti v2.6.0-25-g27e15be) 2.6.1
What steps will reproduce the problem?
# gnt-node modify -D yes $NODE
- master_candidate -> False
- drained -> True
What is the expected output? What do you see instead?
Draining a node makes it no longer a master candidate. If there are too few
master candidates you may need to promote one. Promoting a node to master
candidate can be very slow, if the unarchived job history is large. This is
especially problematic if the purpose of the drain is to do a maintenance
that requires rebooting the whole cluster node by node, because you
basically have to copy the whole job history for each node you reboot. I
think what I'd like is to be able to drain a node for a quick reboot. This
would prevent new instances from being created on the node, prevent
instances from failing over to the node, and other things that would make
it unsafe to reboot the node, but it wouldn't mean the node had to stop
being a master candidate.
If there's another way to make drains faster I'd be open to that, too. I
don't know for sure that this is the right solution.
--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings
