Host blacklist, whitelist and canarilist ( crash if a node is blacklisted AND ( white or canary)) just as an extra safety mechanism for e.g. sql and mgmtnodes
Activity
changed milestone to %Automatic rolling node updates
added Phase1 label
a list of canary nodes for autoupdate. I propose 1 massive node and 1 gpu/desktop node. It could also be a random Idle node of a certain partition. I need two nodes for QA testing with multinode mpi I would suggest one of the m3a nodes + one K1 node if sufficient denyList ( fomerly blacklist ) where I would expect the management and loginnodes as well as certain unicorns like proxy, gt001, syslog, dgx boxes I agree with this allowlist ( formerly whitelist ) . maybe half the short partition or a third of the desktops. The whitelist is intended to grow to all computenodes after a period of trust gaining , and then it should shrink again once we start with Draining and Cold changes We can release some of m3i nodes and you can take 3 nodes in the m3a range and another 3 in the m3f range. This list can be enlarged later as we go