You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@openwhisk.apache.org by Dominic Kim <st...@gmail.com> on 2020/08/20 01:00:38 UTC
Restore the Apache VMs.
Hi Whiskers.
I've been working on restoring Apache VMs to re-enable HA tests.
https://github.com/apache/openwhisk/issues/4939
Now we have one VM available.
I already requested it via Apache Infra ticket, but I am asked to contact
vp-infra@a.o as they are decommissioned once due to low utilization.
https://issues.apache.org/jira/browse/INFRA-20669?focusedCommentId=17178704&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-17178704
I feel like we need to discuss our opinion first before asking him.
Do we really need to get those VMs back?
-dom
Re: Restore the Apache VMs.
Posted by Dominic Kim <st...@gmail.com>.
Yes, I meant the HA tests in the ansible-based deployment.
-dom
2020년 8월 24일 (월) 오전 10:42, Rodric Rabbah <ro...@gmail.com>님이 작성:
> It’s still used in Travis. I think Dave was suggesting kube for Jenkins.
>
> -r
>
> > On Aug 23, 2020, at 9:34 PM, Dominic Kim <st...@gmail.com> wrote:
> >
> > @David
> > Thank you for sharing.
> > I will look into it.
> >
> >> I don't think investing in ansible-based distributed deploy/testing is
> > worth the effort.
> >
> > IMHO, if we would not drop the ansible-based deployment, I think it would
> > still be worth it to guarantee it's working via CI.
> >
> > -dom
> >
> > 2020년 8월 21일 (금) 오후 11:24, David P Grove <gr...@us.ibm.com>님이 작성:
> >
> >>
> >> Hi,
> >>
> >> To answer the question about Jenkins documentation: General
> Apache
> >> Jenkins info:
> https://cwiki.apache.org/confluence/display/INFRA/Jenkins
> >>
> >> Our "normal" Jenkins jobs that (a) build/publish the website on
> >> commits and (b) push nightly controller/invoker images to dockerhub for
> use
> >> by our CI process does not need OpenWhisk-specific VMs. They are
> running
> >> on vanilla "website" and "ubuntu" Jenkins worker nodes shared by all
> >> projects and provided by Infra. The shared nodes are adequate for our
> >> these two jobs.
> >>
> >> We had dedicated nodes that were functional for a short period (a
> >> few
> >> months) to do multi-node testing of PRs for the core repo. This was
> >> triggered by a Jenkins job, but the real work was done via an ansible
> >> deploy/test, so it probably wasn't obvious at the Jenkins level that the
> >> machines were being utilized. This was not a very reliable setup.
> >> Vincent did document the setup in our cwiki (eg
> >>
> https://cwiki.apache.org/confluence/display/OPENWHISK/How+to+maintain+the
> >> +Jenkins+pipeline+for+OpenWhisk), but I would not suggest attempting to
> >> duplicate this on the new ci-builds.a.o Jenkins service.
> >>
> >> My opinion is that we don't really need dedicated Jenkins nodes.
> >> If
> >> we wanted dedicated VMs for testing, we should instead be configuring
> them
> >> as a Kubernetes cluster and doing multi-node testing via a Kubernetes
> >> deployment of OpenWhisk onto those nodes. I don't think investing in
> >> ansible-based distributed deploy/testing is worth the effort. This
> could
> >> be driven by a Jenkins job (which should be able to run on a
> non-dedicated
> >> node as it is doing very little actual work), but it does not need the
> >> Kubernetes cluster to actually be made from Jenkins worker nodes.
> >>
> >> --dave
> >>
>
Re: Restore the Apache VMs.
Posted by Rodric Rabbah <ro...@gmail.com>.
It’s still used in Travis. I think Dave was suggesting kube for Jenkins.
-r
> On Aug 23, 2020, at 9:34 PM, Dominic Kim <st...@gmail.com> wrote:
>
> @David
> Thank you for sharing.
> I will look into it.
>
>> I don't think investing in ansible-based distributed deploy/testing is
> worth the effort.
>
> IMHO, if we would not drop the ansible-based deployment, I think it would
> still be worth it to guarantee it's working via CI.
>
> -dom
>
> 2020년 8월 21일 (금) 오후 11:24, David P Grove <gr...@us.ibm.com>님이 작성:
>
>>
>> Hi,
>>
>> To answer the question about Jenkins documentation: General Apache
>> Jenkins info: https://cwiki.apache.org/confluence/display/INFRA/Jenkins
>>
>> Our "normal" Jenkins jobs that (a) build/publish the website on
>> commits and (b) push nightly controller/invoker images to dockerhub for use
>> by our CI process does not need OpenWhisk-specific VMs. They are running
>> on vanilla "website" and "ubuntu" Jenkins worker nodes shared by all
>> projects and provided by Infra. The shared nodes are adequate for our
>> these two jobs.
>>
>> We had dedicated nodes that were functional for a short period (a
>> few
>> months) to do multi-node testing of PRs for the core repo. This was
>> triggered by a Jenkins job, but the real work was done via an ansible
>> deploy/test, so it probably wasn't obvious at the Jenkins level that the
>> machines were being utilized. This was not a very reliable setup.
>> Vincent did document the setup in our cwiki (eg
>> https://cwiki.apache.org/confluence/display/OPENWHISK/How+to+maintain+the
>> +Jenkins+pipeline+for+OpenWhisk), but I would not suggest attempting to
>> duplicate this on the new ci-builds.a.o Jenkins service.
>>
>> My opinion is that we don't really need dedicated Jenkins nodes.
>> If
>> we wanted dedicated VMs for testing, we should instead be configuring them
>> as a Kubernetes cluster and doing multi-node testing via a Kubernetes
>> deployment of OpenWhisk onto those nodes. I don't think investing in
>> ansible-based distributed deploy/testing is worth the effort. This could
>> be driven by a Jenkins job (which should be able to run on a non-dedicated
>> node as it is doing very little actual work), but it does not need the
>> Kubernetes cluster to actually be made from Jenkins worker nodes.
>>
>> --dave
>>
Re: Restore the Apache VMs.
Posted by Dominic Kim <st...@gmail.com>.
@David
Thank you for sharing.
I will look into it.
> I don't think investing in ansible-based distributed deploy/testing is
worth the effort.
IMHO, if we would not drop the ansible-based deployment, I think it would
still be worth it to guarantee it's working via CI.
-dom
2020년 8월 21일 (금) 오후 11:24, David P Grove <gr...@us.ibm.com>님이 작성:
>
> Hi,
>
> To answer the question about Jenkins documentation: General Apache
> Jenkins info: https://cwiki.apache.org/confluence/display/INFRA/Jenkins
>
> Our "normal" Jenkins jobs that (a) build/publish the website on
> commits and (b) push nightly controller/invoker images to dockerhub for use
> by our CI process does not need OpenWhisk-specific VMs. They are running
> on vanilla "website" and "ubuntu" Jenkins worker nodes shared by all
> projects and provided by Infra. The shared nodes are adequate for our
> these two jobs.
>
> We had dedicated nodes that were functional for a short period (a
> few
> months) to do multi-node testing of PRs for the core repo. This was
> triggered by a Jenkins job, but the real work was done via an ansible
> deploy/test, so it probably wasn't obvious at the Jenkins level that the
> machines were being utilized. This was not a very reliable setup.
> Vincent did document the setup in our cwiki (eg
> https://cwiki.apache.org/confluence/display/OPENWHISK/How+to+maintain+the
> +Jenkins+pipeline+for+OpenWhisk), but I would not suggest attempting to
> duplicate this on the new ci-builds.a.o Jenkins service.
>
> My opinion is that we don't really need dedicated Jenkins nodes.
> If
> we wanted dedicated VMs for testing, we should instead be configuring them
> as a Kubernetes cluster and doing multi-node testing via a Kubernetes
> deployment of OpenWhisk onto those nodes. I don't think investing in
> ansible-based distributed deploy/testing is worth the effort. This could
> be driven by a Jenkins job (which should be able to run on a non-dedicated
> node as it is doing very little actual work), but it does not need the
> Kubernetes cluster to actually be made from Jenkins worker nodes.
>
> --dave
>
RE: Restore the Apache VMs.
Posted by David P Grove <gr...@us.ibm.com>.
Hi,
To answer the question about Jenkins documentation: General Apache
Jenkins info: https://cwiki.apache.org/confluence/display/INFRA/Jenkins
Our "normal" Jenkins jobs that (a) build/publish the website on
commits and (b) push nightly controller/invoker images to dockerhub for use
by our CI process does not need OpenWhisk-specific VMs. They are running
on vanilla "website" and "ubuntu" Jenkins worker nodes shared by all
projects and provided by Infra. The shared nodes are adequate for our
these two jobs.
We had dedicated nodes that were functional for a short period (a few
months) to do multi-node testing of PRs for the core repo. This was
triggered by a Jenkins job, but the real work was done via an ansible
deploy/test, so it probably wasn't obvious at the Jenkins level that the
machines were being utilized. This was not a very reliable setup.
Vincent did document the setup in our cwiki (eg
https://cwiki.apache.org/confluence/display/OPENWHISK/How+to+maintain+the
+Jenkins+pipeline+for+OpenWhisk), but I would not suggest attempting to
duplicate this on the new ci-builds.a.o Jenkins service.
My opinion is that we don't really need dedicated Jenkins nodes. If
we wanted dedicated VMs for testing, we should instead be configuring them
as a Kubernetes cluster and doing multi-node testing via a Kubernetes
deployment of OpenWhisk onto those nodes. I don't think investing in
ansible-based distributed deploy/testing is worth the effort. This could
be driven by a Jenkins job (which should be able to run on a non-dedicated
node as it is doing very little actual work), but it does not need the
Kubernetes cluster to actually be made from Jenkins worker nodes.
--dave
Re: Restore the Apache VMs.
Posted by Dominic Kim <st...@gmail.com>.
Frankly speaking, I have no idea about the Jenkins and its slaves.
I thought we can get VMs via Infra ticket only. Is there any automated way
to get VMs for Jenkins?
It would be also thankful if you direct me to any documents about how our
Jenkins CI works(had worked).
Thanks.
-dom
2020년 8월 21일 (금) 오전 4:36, Rodric Rabbah <ro...@gmail.com>님이 작성:
> We needed those VMs to setup dedicated multi-invoker deployments. It is
> true that at some point the jk build stopped working.
> Is the issue now that we can't get any VMs at all in Jenkins if we don't
> restore these VMs?
>
> -r
>
> On Wed, Aug 19, 2020 at 9:01 PM Dominic Kim <st...@gmail.com> wrote:
>
> > Hi Whiskers.
> >
> > I've been working on restoring Apache VMs to re-enable HA tests.
> > https://github.com/apache/openwhisk/issues/4939
> >
> > Now we have one VM available.
> >
> > I already requested it via Apache Infra ticket, but I am asked to contact
> > vp-infra@a.o as they are decommissioned once due to low utilization.
> >
> >
> https://issues.apache.org/jira/browse/INFRA-20669?focusedCommentId=17178704&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-17178704
> >
> > I feel like we need to discuss our opinion first before asking him.
> > Do we really need to get those VMs back?
> >
> > -dom
> >
>
Re: Restore the Apache VMs.
Posted by Rodric Rabbah <ro...@gmail.com>.
We needed those VMs to setup dedicated multi-invoker deployments. It is
true that at some point the jk build stopped working.
Is the issue now that we can't get any VMs at all in Jenkins if we don't
restore these VMs?
-r
On Wed, Aug 19, 2020 at 9:01 PM Dominic Kim <st...@gmail.com> wrote:
> Hi Whiskers.
>
> I've been working on restoring Apache VMs to re-enable HA tests.
> https://github.com/apache/openwhisk/issues/4939
>
> Now we have one VM available.
>
> I already requested it via Apache Infra ticket, but I am asked to contact
> vp-infra@a.o as they are decommissioned once due to low utilization.
>
> https://issues.apache.org/jira/browse/INFRA-20669?focusedCommentId=17178704&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-17178704
>
> I feel like we need to discuss our opinion first before asking him.
> Do we really need to get those VMs back?
>
> -dom
>