ovirt: Use the DNSVIP for workers #1313
Conversation
openshift-ci-robot
added
the
size/M
|
💯 to removing templating where possible. I only did a quick look over and don't see anything jump out at me. Lemme go through it more thoughtfully to be sure of the underlying logic. lmk if you need this merged more urgently tho. |
|
@ericavonb do note that the first commit here is taken from a PR #1298 |
|
/retest |
1 similar comment
|
/retest |
|
@rgolangh Any reason for including the static pods resources customizations to this PR? |
|
That was just to help me run them without failing on memory. In the end it
should be removed. (it has a separate PR)
…On Thu, 12 Dec 2019 at 15:45, Yossi Boaron ***@***.***> wrote:
@rgolangh <https://github.com/rgolangh> Any reason for including the
static pods resources customizations to this PR?
I would suggest splitting this PR
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#1313?email_source=notifications&email_token=ABGBYHBBRPUXZB4ZJKRJQJDQYI553A5CNFSM4JX44JZKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEGWWQ4Y#issuecomment-565012595>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ABGBYHDDWZCMXTMCWABCFMLQYI553ANCNFSM4JX44JZA>
.
|
|
@ericavonb I can't seem to get rid of the aws errors. Is that a known issue? |
|
/retest |
Looks like AWS has not doing well yesterday: https://prow.svc.ci.openshift.org/job-history/origin-ci-test/pr-logs/directory/pull-ci-openshift-machine-config-operator-master-e2e-aws?buildId= I'll look at this PR's failures to make sure they are failing for the same reason the others were. |
|
Looks like e2e-aws passed 🎉 ! Leaving the lgmt to someone from the ovirt side. /approve |
openshift-ci-robot
added
the
approved
The way we provision the non-virtual IP nameserver is not working in a reliable way for workers. Instead use the known to work way of using the DNSVIP in dhclient.conf which is written directly by ignition to disk, so NM picks is up with no known issues when the machine starts. Signed-off-by: Roy Golan <rgolan@redhat.com>
rgolangh
force-pushed
the
use_dnsvip_for_workers
branch
from
ed65df3 to
a4a4e92
Compare
|
/lgtm |
|
@rgolangh: you cannot LGTM your own PR. DetailsIn response to this:
Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository. |
|
/retest |
1 similar comment
|
/retest |
|
@ericavonb looks we are ready now. Can this be approved by somebody else on the team? |
|
/lgtm |
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: celebdor, ericavonb, rgolangh The full list of commands accepted by this bot can be found here. The pull request process is described here DetailsNeeds approval from an approver in each of these files:
Approvers can indicate their approval by writing |
7 participants
oVirt worker will most of the time failed to join the cluster because they don't have a nameserver
set to resolve the api or api-int.
The problem is that a
/etc/dhclient/dhclient.confis written during the network manager pre-up hook, using the file d3d4e5f#diff-35faebd7db8179e62b9eb7ca9604ce72and because of a race, dhclient doesn't recognize the new config.
This change will write
/etc/dhcp/dhclient.confusing ignition, with the DNS VIP prependedas the first nameserver. The DNSVIP is interpolated by the server so it is written in its final form
in the ignition phase (just like the masters). This eliminates the race I'm seeing.
Related-to: #1312
Run IPI installation of oVirt, 3 master (all of them use those static pods) and 2 workers which are added using cluster-api-provider-ovirt, and see them join and go into
Readystate.Set the
DNSVIPas the first nameserver for worker by writing/etc/dhcp/dhclient.confusing ignition.