CLOUDSTACK-9362: Skip VXLANs when rewriting the bridge name for migrations (4.8-2) #1513
Conversation
|
"Richard Klein (RSI)" on dev@cloudstack.apache.org replies: We worked around the issue by just changing the name of the bridge from brv= Richard Klein=A0 rklein@rsitex.com=20 |
|
We'll pull this in for testing. |
|
Tested this in a hardware 4.8 lab: I was able to migrate VXLAN enabled VMs between hosts as expected. LGTM |
|
Also, tested in a 4.8 lab and I was able to migrate VMs between hosts when using VXLAN Guest Isolated network. LGTM |
CI RESULTSSummary of the problem(s): Associated Uploads
Uploads will be available until Comment created by |
|
@insom, @dmabry and @kiwiflyer, the CI failure seems like it could be related to this change, but I can't be sure. Can you guys provide some guidance since all of you have tested this. Also, @insom can you do a force push or close and re-open the PR so we can get this green to make it mergeable once we agree on status. Thanks... |
|
@swill I've closed and re-opened. This change is to a shell script only and the failures are within Marvin, so I'm pretty confident that they are unrelated to my proposed change. edit: As in, they are errors in the simulated DC, which I don't believe touches these scripts, not Marvin itself. |
|
@insom yes, I tend to agree with you. the main reason I highlighted it was because the error was network related and the code changed details related to an interface, so I figured I would check. Once everything comes back green, I will merge this. Thank you for the work and the quick reply. 👍 |
|
@swill Hmm. Now it's a different Cobertura error |
|
Holy errors batman!!! It looks like it started like this: I am pretty sure we are having clean problems on Jenkins, so I wonder if it is related to something being left over from the build of a different branch. Hmmm, I need to get better with Jenkins. Does anyone have an expertise here that could potentially help us understand what would be happening here? |
|
@insom we are down to the wire here. I would like to see if we can get this green and merge it before the freeze. Can you rebase and push again to kick the automation. I will likely make judgement calls and merge PRs that have a passing travis even if jenkins is failing. We will see. I would rather not be in that situation... |
|
@swill rebased and pushed. It does have two LGTMs and tests by real people (plus it's in production at here at iWeb) if that makes it a little easier to overrule Jenkins. |
|
We are green now. 👍 I am going to rerun CI to make sure the Failure we got does not pop up again. Otherwise we are pretty good. Thanks... |
CI RESULTSSummary of the problem(s): Associated Uploads
Uploads will be available until Comment created by |
|
Not sure what to think of that. I have never seen those errors before in my environment... |
|
@swill Sure you haven't got some artifacts left behind from some other testing? This is only a libvirt hook, so I can't imagine how the tests failing above could be related to this PR. |
|
I will rerun tests now... |
CI RESULTSAssociated Uploads
Uploads will be available until Comment created by |
|
The tests that were failing are not failing anymore, I think this is ready... |
CLOUDSTACK-9362: Skip VXLANs when rewriting the bridge name for migrations (4.8-2)From the [JIRA issue](https://issues.apache.org/jira/browse/CLOUDSTACK-9362): > bb8f7c6 > > The above commit introduces rewriting of bridge device names when migrating a virtual machine from one host to another. However, it also matches bridges called "brvx-1234" and rewrites them to (in my case) "brem1-1234" - this doesn't match the bridge name on the destination and causes the migration to fail with the error: > > error : virNetDevGetMTU:397 : Cannot get interface MTU on 'brem1-1234': No such device > > I have flagged this as major because it's not possible to migrate VMs using VXLANs for maintenance, which seems important (it's certainly important to me!). This is a version of #1508 based against 4.8 (sorry!) * pr/1513: Skip VXLANs when rewriting the bridge name for migrations Signed-off-by: Will Stevens <williamstevens@gmail.com>
6 participants
From the JIRA issue:
This is a version of #1508 based against 4.8 (sorry!)