Hostname patch for vsphere provider limitations with juju #44780
Conversation
k8s-ci-robot
added
the
cncf-cla: yes
|
This change is |
k8s-github-robot
added
approved
| MY_HOSTNAME=$(hostname) | ||
|
|
||
| if [ "${MY_HOSTNAME}" == "ubuntuguest" ]; then | ||
| echo "Detected broken vsphere integration. Applying hostname override" |
|
|
||
| if [ "${MY_HOSTNAME}" == "ubuntuguest" ]; then | ||
| echo "Detected broken vsphere integration. Applying hostname override" | ||
| if [ -z "${JUJU_UNIT_NAME}" ]; then |
There was a problem hiding this comment.
We should do bash fallback instead of just exiting. For example
: ${JUJU_UNIT_NAME:=`uuidgen`}
At the top of the script will make sure that JUJU_UNIT_NAME exists or is assigned to a UUID
There was a problem hiding this comment.
Ok, We can certainly do that. +1
|
I think this is a good place to resolve this, in exec.d, just a few places to simplify logic |
lazypower
force-pushed
the
vsphere-hostname-patch
branch
from
2adc196
to
03bf672
Compare
This patch sets the hostname to a unique identifier (the juju unit name) during pre-deployment of the charm. This may not be a FQDN resolveable hostname but will prevent hostname collision.
lazypower
force-pushed
the
vsphere-hostname-patch
branch
from
03bf672
to
32acffc
Compare
|
@marcoceppi i took your suggestions wholesale and verified they work as intended. Ready for another pass on this one. |
|
@k8s-bot gce etcd3 e2e test this |
|
/assign @marcoceppi |
|
/lgtm |
k8s-ci-robot
added
the
lgtm
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: chuckbutler, marcoceppi
Needs approval from an approver in each of these OWNERS Files:
You can indicate your approval by writing |
|
Automatic merge from submit-queue (batch tested with PRs 42486, 44780) |
What this PR does / why we need it:
The Juju VSphere provider doesn't set a unique hostname which causes issues when scaling worker-pools and they all have the hostname
ubuntuguest. Instead we assign the JUJU_UNIT_NAME to that hostname to prevent the collision which allows the master to sort out that there are multiple units and not one attempting re-registration.Which issue this PR fixes (optional, in
fixes #<issue number>(, fixes #<issue_number>, ...)format, will close that issue when PR gets merged): fixes https://github.com/juju-solutions/bundle-canonical-kubernetes/issues/237Special notes for your reviewer:
The charm-pre-exec runs before it installs the charm software so the validation can happen quickly. Check hostname output, as well as kubectl get no post deployment.