workaround for the lack of native el8 slaves in the CI
Hi all, we're trying to move vdsm tests to el8. AFAIK we don't have native el8 slaves in the CI yet, so our el8 stages run on el7 with el8 mocked. There are some issues with scripts setting up the env for storage tests - they are refusing to work on older kernel versions. I have an idea for a workaround - until now we were using 'host-distro: same' in our 'stdci.yaml'. I can see that this option can also take a value of 'better/newer' [1], but that won't work for el8. Could we possibly implement something like [2] ? That would allow us to i.e. define an el8 mocked env with 'host-distro: fc30' enforcing native fc30 host. I assume this would be a temporary workaround only, until we have native el8 slaves. Regards, Marcin [1] https://github.com/oVirt/jenkins/blob/a796094bfd15f1a85b9f630aaeb25ce6e0dab3... [2] https://gerrit.ovirt.org/#/c/104091/
On středa 16. října 2019 10:53:45 CEST Marcin Sobczyk wrote:
Hi all,
we're trying to move vdsm tests to el8.
AFAIK we don't have native el8 slaves in the CI yet, so our el8 stages run on el7 with el8 mocked. There are some issues with scripts setting up the env for storage tests - they are refusing to work on older kernel versions.
I have an idea for a workaround - until now we were using 'host-distro: same' in our 'stdci.yaml'. I can see that this option can also take a value of 'better/newer' [1], but that won't work for el8. Could we possibly implement something like [2] ? That would allow us to i.e. define an el8 mocked env with 'host-distro: fc30' enforcing native fc30 host. I assume this would be a temporary workaround only, until we have native el8 slaves.
if it works, it would be great
Regards, Marcin
[1] https://github.com/oVirt/jenkins/blob/a796094bfd15f1a85b9f630aaeb25ce6e0dab3 5d/pipelines/libs/stdci_runner.groovy#L127 [2] https://gerrit.ovirt.org/#/c/104091/
On Wed, 16 Oct 2019 at 12:21, Vojtech Juranek <vjuranek@redhat.com> wrote:
On středa 16. října 2019 10:53:45 CEST Marcin Sobczyk wrote:
Hi all,
we're trying to move vdsm tests to el8.
AFAIK we don't have native el8 slaves in the CI yet, so our el8 stages run on el7 with el8 mocked. There are some issues with scripts setting up the env for storage tests - they are refusing to work on older kernel versions.
I have an idea for a workaround - until now we were using 'host-distro: same' in our 'stdci.yaml'. I can see that this option can also take a value of 'better/newer' [1], but that won't work for el8. Could we possibly implement something like [2] ? That would allow us to i.e. define an el8 mocked env with 'host-distro: fc30' enforcing native fc30 host. I assume this would be a temporary workaround only, until we have native el8 slaves.
if it works, it would be great
Given the rapid pace in which Fedora versions go EOL, I wouldn't want to support this kind of syntax. Instead, I suggest we just make sure that `newer` means fc30+ for el8. We're not really planning to ever enable EL8 slaves, since there is a huge amount of issues we'll need to solve and test to make that happen, and we're looking to deprecate the static way in which we currently manage slaves. In other words, any workaround you implement now, may not be as temporary as you think.
Regards, Marcin
[1]
https://github.com/oVirt/jenkins/blob/a796094bfd15f1a85b9f630aaeb25ce6e0dab3
5d/pipelines/libs/stdci_runner.groovy#L127 [2] https://gerrit.ovirt.org/#/c/104091/
_______________________________________________ Devel mailing list -- devel@ovirt.org To unsubscribe send an email to devel-leave@ovirt.org Privacy Statement: https://www.ovirt.org/site/privacy-policy/ oVirt Code of Conduct: https://www.ovirt.org/community/about/community-guidelines/ List Archives: https://lists.ovirt.org/archives/list/devel@ovirt.org/message/V765YTTZLZV4O7...
-- Barak Korren RHV DevOps team , RHCE, RHCi Red Hat EMEA redhat.com | TRIED. TESTED. TRUSTED. | redhat.com/trusted
On Tue, Oct 22, 2019 at 1:50 PM Barak Korren <bkorren@redhat.com> wrote:
On Wed, 16 Oct 2019 at 12:21, Vojtech Juranek <vjuranek@redhat.com> wrote:
Hi all,
we're trying to move vdsm tests to el8.
AFAIK we don't have native el8 slaves in the CI yet, so our el8 stages run on el7 with el8 mocked. There are some issues with scripts setting up the env for storage tests
On středa 16. října 2019 10:53:45 CEST Marcin Sobczyk wrote: -
they are refusing to work on older kernel versions.
I have an idea for a workaround - until now we were using 'host-distro: same' in our 'stdci.yaml'. I can see that this option can also take a value of 'better/newer' [1], but that won't work for el8. Could we possibly implement something like [2] ? That would allow us to i.e. define an el8 mocked env with 'host-distro: fc30' enforcing native fc30 host. I assume this would be a temporary workaround only, until we have native el8 slaves.
if it works, it would be great
Given the rapid pace in which Fedora versions go EOL, I wouldn't want to support this kind of syntax. Instead, I suggest we just make sure that `newer` means fc30+ for el8.
Barak, how soon can we apply this workaround so we can unblock CI for projects?
We're not really planning to ever enable EL8 slaves, since there is a huge amount of issues we'll need to solve and test to make that happen, and we're looking to deprecate the static way in which we currently manage slaves. In other words, any workaround you implement now, may not be as temporary as you think.
Regards, Marcin
[1]
https://github.com/oVirt/jenkins/blob/a796094bfd15f1a85b9f630aaeb25ce6e0dab3
5d/pipelines/libs/stdci_runner.groovy#L127 [2] https://gerrit.ovirt.org/#/c/104091/
_______________________________________________ Devel mailing list -- devel@ovirt.org To unsubscribe send an email to devel-leave@ovirt.org Privacy Statement: https://www.ovirt.org/site/privacy-policy/ oVirt Code of Conduct: https://www.ovirt.org/community/about/community-guidelines/ List Archives: https://lists.ovirt.org/archives/list/devel@ovirt.org/message/V765YTTZLZV4O7...
-- Barak Korren RHV DevOps team , RHCE, RHCi Red Hat EMEA redhat.com | TRIED. TESTED. TRUSTED. | redhat.com/trusted _______________________________________________ Devel mailing list -- devel@ovirt.org To unsubscribe send an email to devel-leave@ovirt.org Privacy Statement: https://www.ovirt.org/site/privacy-policy/ oVirt Code of Conduct: https://www.ovirt.org/community/about/community-guidelines/ List Archives: https://lists.ovirt.org/archives/list/devel@ovirt.org/message/EDVX2VCPEHDSL2...
-- Eyal edri He / Him / His MANAGER CONTINUOUS PRODUCTIZATION SYSTEM ENGINEERING Red Hat <https://www.redhat.com/> <https://red.ht/sig> phone: +972-9-7692018 irc: eedri (on #tlv #rhev-dev #rhev-integ #cp-devel)
On Tue, 22 Oct 2019 at 13:54, Eyal Edri <eedri@redhat.com> wrote:
On Tue, Oct 22, 2019 at 1:50 PM Barak Korren <bkorren@redhat.com> wrote:
On Wed, 16 Oct 2019 at 12:21, Vojtech Juranek <vjuranek@redhat.com> wrote:
On středa 16. října 2019 10:53:45 CEST Marcin Sobczyk wrote:
Hi all,
we're trying to move vdsm tests to el8.
AFAIK we don't have native el8 slaves in the CI yet, so our el8 stages run on el7 with el8 mocked. There are some issues with scripts setting up the env for storage tests - they are refusing to work on older kernel versions.
I have an idea for a workaround - until now we were using 'host-distro: same' in our 'stdci.yaml'. I can see that this option can also take a value of 'better/newer' [1], but that won't work for el8. Could we possibly implement something like [2] ? That would allow us to i.e. define an el8 mocked env with 'host-distro: fc30' enforcing native fc30 host. I assume this would be a temporary workaround only, until we have native el8 slaves.
if it works, it would be great
Given the rapid pace in which Fedora versions go EOL, I wouldn't want to support this kind of syntax. Instead, I suggest we just make sure that `newer` means fc30+ for el8.
Barak, how soon can we apply this workaround so we can unblock CI for projects?
Here is a patch that does that: https://gerrit.ovirt.org/c/104165/ Now its just a matter of reviews
We're not really planning to ever enable EL8 slaves, since there is a huge amount of issues we'll need to solve and test to make that happen, and we're looking to deprecate the static way in which we currently manage slaves. In other words, any workaround you implement now, may not be as temporary as you think.
Regards, Marcin
[1]
https://github.com/oVirt/jenkins/blob/a796094bfd15f1a85b9f630aaeb25ce6e0dab3
5d/pipelines/libs/stdci_runner.groovy#L127 [2] https://gerrit.ovirt.org/#/c/104091/
_______________________________________________ Devel mailing list -- devel@ovirt.org To unsubscribe send an email to devel-leave@ovirt.org Privacy Statement: https://www.ovirt.org/site/privacy-policy/ oVirt Code of Conduct: https://www.ovirt.org/community/about/community-guidelines/ List Archives: https://lists.ovirt.org/archives/list/devel@ovirt.org/message/V765YTTZLZV4O7...
-- Barak Korren RHV DevOps team , RHCE, RHCi Red Hat EMEA redhat.com | TRIED. TESTED. TRUSTED. | redhat.com/trusted _______________________________________________ Devel mailing list -- devel@ovirt.org To unsubscribe send an email to devel-leave@ovirt.org Privacy Statement: https://www.ovirt.org/site/privacy-policy/ oVirt Code of Conduct: https://www.ovirt.org/community/about/community-guidelines/ List Archives: https://lists.ovirt.org/archives/list/devel@ovirt.org/message/EDVX2VCPEHDSL2...
--
Eyal edri
He / Him / His
MANAGER
CONTINUOUS PRODUCTIZATION
SYSTEM ENGINEERING
Red Hat <https://www.redhat.com/> <https://red.ht/sig> phone: +972-9-7692018 irc: eedri (on #tlv #rhev-dev #rhev-integ #cp-devel)
-- Barak Korren RHV DevOps team , RHCE, RHCi Red Hat EMEA redhat.com | TRIED. TESTED. TRUSTED. | redhat.com/trusted
On Wed, Oct 16, 2019 at 11:53 AM Marcin Sobczyk <msobczyk@redhat.com> wrote:
Hi all,
we're trying to move vdsm tests to el8.
AFAIK we don't have native el8 slaves in the CI yet, so our el8 stages run on el7 with el8 mocked. There are some issues with scripts setting up the env for storage tests - they are refusing to work on older kernel versions.
This is true, but the issue is only xfs. We can enable the tests by switching to ext4 temporarily. We can revert this: commit b520d0df06129dd9d8d7f76f08425a3007962ab5 Author: Nir Soffer <nsoffer@redhat.com> Date: Tue Aug 27 18:32:19 2019 +0300 tests: Revert "tests: use ext4 for userstorage" Then moving to https://github.com/nirs/userstorage will make it possible to have an optional xfs backend that will be skipped only on Jenkins, but will run the xfs tests on travis and when running locally. This is what we do now in imageio: https://gerrit.ovirt.org/c/103958/ We can enable xfs again when we have a way to run el8 slaves on fedora >= 29. I have an idea for a workaround - until now we were using 'host-distro:
same' in our 'stdci.yaml'. I can see that this option can also take a value of 'better/newer' [1], but that won't work for el8. Could we possibly implement something like [2] ? That would allow us to i.e. define an el8 mocked env with 'host-distro: fc30' enforcing native fc30 host. I assume this would be a temporary workaround only, until we have native el8 slaves.
Regards, Marcin
[1] https://github.com/oVirt/jenkins/blob/a796094bfd15f1a85b9f630aaeb25ce6e0dab3... [2] https://gerrit.ovirt.org/#/c/104091/
On 22 Oct 2019, at 14:04, Nir Soffer <nsoffer@redhat.com> wrote:
On Wed, Oct 16, 2019 at 11:53 AM Marcin Sobczyk <msobczyk@redhat.com <mailto:msobczyk@redhat.com>> wrote: Hi all,
we're trying to move vdsm tests to el8.
AFAIK we don't have native el8 slaves in the CI yet, so our el8 stages run on el7 with el8 mocked. There are some issues with scripts setting up the env for storage tests - they are refusing to work on older kernel versions.
This is true, but the issue is only xfs. We can enable the tests by switching to ext4 temporarily.
We can revert this:
commit b520d0df06129dd9d8d7f76f08425a3007962ab5 Author: Nir Soffer <nsoffer@redhat.com <mailto:nsoffer@redhat.com>> Date: Tue Aug 27 18:32:19 2019 +0300
tests: Revert "tests: use ext4 for userstorage"
Then moving to https://github.com/nirs/userstorage <https://github.com/nirs/userstorage> will make it possible to have an optional xfs backend that will be skipped only on Jenkins, but will run the xfs tests on travis and when running locally. This is what we do now in imageio: https://gerrit.ovirt.org/c/103958/ <https://gerrit.ovirt.org/c/103958/>
We can enable xfs again when we have a way to run el8 slaves on fedora >= 29.
That sounds great, can you go for it? (at least the first part, to unblock the jenkins CI )
I have an idea for a workaround - until now we were using 'host-distro: same' in our 'stdci.yaml'. I can see that this option can also take a value of 'better/newer' [1], but that won't work for el8. Could we possibly implement something like [2] ? That would allow us to i.e. define an el8 mocked env with 'host-distro: fc30' enforcing native fc30 host. I assume this would be a temporary workaround only, until we have native el8 slaves.
Regards, Marcin
[1] https://github.com/oVirt/jenkins/blob/a796094bfd15f1a85b9f630aaeb25ce6e0dab3... <https://github.com/oVirt/jenkins/blob/a796094bfd15f1a85b9f630aaeb25ce6e0dab35d/pipelines/libs/stdci_runner.groovy#L127> [2] https://gerrit.ovirt.org/#/c/104091/ <https://gerrit.ovirt.org/#/c/104091/>
On Wed, 23 Oct 2019 at 16:28, Michal Skrivanek <michal.skrivanek@redhat.com> wrote:
On 22 Oct 2019, at 14:04, Nir Soffer <nsoffer@redhat.com> wrote:
On Wed, Oct 16, 2019 at 11:53 AM Marcin Sobczyk <msobczyk@redhat.com> wrote:
Hi all,
we're trying to move vdsm tests to el8.
AFAIK we don't have native el8 slaves in the CI yet, so our el8 stages run on el7 with el8 mocked. There are some issues with scripts setting up the env for storage tests - they are refusing to work on older kernel versions.
This is true, but the issue is only xfs. We can enable the tests by switching to ext4 temporarily.
We can revert this:
commit b520d0df06129dd9d8d7f76f08425a3007962ab5 Author: Nir Soffer <nsoffer@redhat.com> Date: Tue Aug 27 18:32:19 2019 +0300
tests: Revert "tests: use ext4 for userstorage"
Then moving to https://github.com/nirs/userstorage will make it possible to have an optional xfs backend that will be skipped only on Jenkins, but will run the xfs tests on travis and when running locally. This is what we do now in imageio: https://gerrit.ovirt.org/c/103958/
We can enable xfs again when we have a way to run el8 slaves on fedora >= 29.
That sounds great, can you go for it? (at least the first part, to unblock the jenkins CI )
Reminder that we have a one-liner patch that would probably unblock things as well: - https://gerrit.ovirt.org/c/104165/
I have an idea for a workaround - until now we were using 'host-distro:
same' in our 'stdci.yaml'. I can see that this option can also take a value of 'better/newer' [1], but that won't work for el8. Could we possibly implement something like [2] ? That would allow us to i.e. define an el8 mocked env with 'host-distro: fc30' enforcing native fc30 host. I assume this would be a temporary workaround only, until we have native el8 slaves.
Regards, Marcin
[1] https://github.com/oVirt/jenkins/blob/a796094bfd15f1a85b9f630aaeb25ce6e0dab3... [2] https://gerrit.ovirt.org/#/c/104091/
-- Barak Korren RHV DevOps team , RHCE, RHCi Red Hat EMEA redhat.com | TRIED. TESTED. TRUSTED. | redhat.com/trusted
participants (6)
-
Barak Korren -
Eyal Edri -
Marcin Sobczyk -
Michal Skrivanek -
Nir Soffer -
Vojtech Juranek