Only schedule to pods that are available. This turns on the node #5725
Conversation
|
Thanks for your pull request. It looks like this may be your first contribution to a Google open source project, in which case you'll need to sign a Contributor License Agreement (CLA) at https://cla.developers.google.com/. If you've already signed a CLA, it's possible we don't have your GitHub username or you're using a different email address. Check the information on your CLA or see this help article on setting the email on your git commits. Once you've done that, please reply here to let us know. If you signed the CLA as a corporation, please let us know the company's name. |
rrati
force-pushed
the
schedule-to-available-nodes-5545
branch
2 times, most recently
from
e127217 to
4c4ae9b
Compare
|
@ddysher Any chance you can take a look? |
| @@ -77,7 +77,7 @@ func NewCMServer() *CMServer { | |||
| NodeMilliCPU: 1000, | |||
| NodeMemory: resource.MustParse("3Gi"), | |||
| SyncNodeList: true, | |||
| SyncNodeStatus: false, | |||
| SyncNodeStatus: true, | |||
There was a problem hiding this comment.
why change the default value? we are not syncing node status in controller manager now.
There was a problem hiding this comment.
In order to prevent scheduling to nodes that can not run pods there needs to be a constant monitoring and updating of all node statuses. If the node goes down, gets rebooted, or even re-purposed then that node isn't available to run pods. The only way that I see in the current implementation where that is noticed is with the node controller pinging the node and updating the status for the scheduler.
A better way would probably be for the apiserver to require an update only be so old for it to be valid.
There was a problem hiding this comment.
See #5399, which should address you concern. Enabling sync status will mess up the cluster, since it operates in two different modes (node controller pull status vs kubelet push status)
There was a problem hiding this comment.
Yep, that will do exactly what I was trying to accomplish. I'll re-push and remove the node controller updating change
rrati
force-pushed
the
schedule-to-available-nodes-5545
branch
from
4c4ae9b to
cbf0541
Compare
rrati
force-pushed
the
schedule-to-available-nodes-5545
branch
from
cbf0541 to
c2938b2
Compare
|
LGTM |
|
LGTM, thx |
ddysher
added
lgtm
Only schedule to pods that are available. This turns on the node
#5545