[v0.16.0] make flannel subnetLen configurable for big clusters #1848
Conversation
k8s-ci-robot
added
the
cncf-cla: yes
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: The full list of commands accepted by this bot can be found here.
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
k8s-ci-robot
added
the
size/S
Fsero
force-pushed
the
make_flannel_leases_configurable
branch
from
edf6ec1
to
24492f7
Compare
Codecov Report
@@ Coverage Diff @@
## master #1848 +/- ##
==========================================
- Coverage 24.41% 24.40% -0.01%
==========================================
Files 98 98
Lines 5116 5117 +1
==========================================
Hits 1249 1249
- Misses 3727 3728 +1
Partials 140 140
Continue to review full report at Codecov.
|
|
This is awesome, we have the same issue! Thank you. |
|
/lgtm |
dominicgunn
changed the title
We had an issue in a big cluster (around 200 nodes) when rolling upgrade a change and replacing instances we ran out of subnet leases, that causes that new workers were unable to get a flannel lease and hence setting up the network and were killed due to not sending the cfn-signal in time.
This patch aims to make subnetLen configurable for big clusters, using kube-aws default flannel config the maximum is 255 nodes which means around 127 nodes and a rolling upgrade that replaces all instances (like updating the base AMI).
In the case of being unset, flanneld will keep using the default which is /24 per node.