Greetings.
image composes using oz take a long time on our s390x image builder, (due to them being in another datacenter). They take so long, they timeout.
See: https://pagure.io/koji/pull-request/837 from sinny and https://pagure.io/koji/pull-request/841 which makes this value configurable in kojid.conf
I would like to:
1. Update koji on buildvm-s390x-02 to the version I just built in rawhide/f28 that has a backport of the patch that makes this value configurable.
2. Apply the following ansible patch to set the timeout only on the buildvm-s390x-02 builder:
diff --git a/roles/koji_builder/templates/kojid.conf b/roles/koji_builder/templates/kojid.conf index 832d80c..9f10d9c 100644 --- a/roles/koji_builder/templates/kojid.conf +++ b/roles/koji_builder/templates/kojid.conf @@ -20,7 +20,12 @@ maxjobs=25 keepalive=False
rpmbuild_timeout=172800 - +{% if ansible_hostname.startswith('buildvm-s390x-02') %} +; Set oz timeout higher on x390x image builder to allow it to finish. +; Install timeout(seconds) for image build +; if it's unset, use the number in /etc/oz/oz.cfg, supported since oz-0.16.0 +oz_install_timeout=14400 +{% endif %} use_createrepo_c=True
{% if host in groups['buildvm-s390x'] %}
3. Run the playbook on it to update the config and restart kojid.
+1s?
kevin
+1. Do we know 14400 will be enough?
-re
On Fri, Mar 16, 2018 at 3:29 PM, Kevin Fenzi kevin@scrye.com wrote:
Greetings.
image composes using oz take a long time on our s390x image builder, (due to them being in another datacenter). They take so long, they timeout.
See: https://pagure.io/koji/pull-request/837 from sinny and https://pagure.io/koji/pull-request/841 which makes this value configurable in kojid.conf
I would like to:
- Update koji on buildvm-s390x-02 to the version I just built in
rawhide/f28 that has a backport of the patch that makes this value configurable.
- Apply the following ansible patch to set the timeout only on the
buildvm-s390x-02 builder:
diff --git a/roles/koji_builder/templates/kojid.conf b/roles/koji_builder/templates/kojid.conf index 832d80c..9f10d9c 100644 --- a/roles/koji_builder/templates/kojid.conf +++ b/roles/koji_builder/templates/kojid.conf @@ -20,7 +20,12 @@ maxjobs=25 keepalive=False
rpmbuild_timeout=172800
+{% if ansible_hostname.startswith('buildvm-s390x-02') %} +; Set oz timeout higher on x390x image builder to allow it to finish. +; Install timeout(seconds) for image build +; if it's unset, use the number in /etc/oz/oz.cfg, supported since oz-0.16.0 +oz_install_timeout=14400 +{% endif %} use_createrepo_c=True
{% if host in groups['buildvm-s390x'] %}
- Run the playbook on it to update the config and restart kojid.
+1s?
kevin
infrastructure mailing list -- infrastructure@lists.fedoraproject.org To unsubscribe send an email to infrastructure-leave@lists.fedoraproject.org
On 03/16/2018 12:42 PM, Ricky Elrod wrote:
+1. Do we know 14400 will be enough?
We don't, but I figured we could start there and adjust it if it wasn't enough still.
kevin --
-re
On Fri, Mar 16, 2018 at 3:29 PM, Kevin Fenzi kevin@scrye.com wrote:
Greetings.
image composes using oz take a long time on our s390x image builder, (due to them being in another datacenter). They take so long, they timeout.
See: https://pagure.io/koji/pull-request/837 from sinny and https://pagure.io/koji/pull-request/841 which makes this value configurable in kojid.conf
I would like to:
- Update koji on buildvm-s390x-02 to the version I just built in
rawhide/f28 that has a backport of the patch that makes this value configurable.
- Apply the following ansible patch to set the timeout only on the
buildvm-s390x-02 builder:
diff --git a/roles/koji_builder/templates/kojid.conf b/roles/koji_builder/templates/kojid.conf index 832d80c..9f10d9c 100644 --- a/roles/koji_builder/templates/kojid.conf +++ b/roles/koji_builder/templates/kojid.conf @@ -20,7 +20,12 @@ maxjobs=25 keepalive=False
rpmbuild_timeout=172800
+{% if ansible_hostname.startswith('buildvm-s390x-02') %} +; Set oz timeout higher on x390x image builder to allow it to finish. +; Install timeout(seconds) for image build +; if it's unset, use the number in /etc/oz/oz.cfg, supported since oz-0.16.0 +oz_install_timeout=14400 +{% endif %} use_createrepo_c=True
{% if host in groups['buildvm-s390x'] %}
- Run the playbook on it to update the config and restart kojid.
+1s?
kevin
infrastructure mailing list -- infrastructure@lists.fedoraproject.org To unsubscribe send an email to infrastructure-leave@lists.fedoraproject.org
infrastructure mailing list -- infrastructure@lists.fedoraproject.org To unsubscribe send an email to infrastructure-leave@lists.fedoraproject.org
+1 to try
On Fri, Mar 16, 2018 at 8:51 PM, Kevin Fenzi kevin@scrye.com wrote:
On 03/16/2018 12:42 PM, Ricky Elrod wrote:
+1. Do we know 14400 will be enough?
We don't, but I figured we could start there and adjust it if it wasn't enough still.
kevin
-re
On Fri, Mar 16, 2018 at 3:29 PM, Kevin Fenzi kevin@scrye.com wrote:
Greetings.
image composes using oz take a long time on our s390x image builder, (due to them being in another datacenter). They take so long, they timeout.
See: https://pagure.io/koji/pull-request/837 from sinny and https://pagure.io/koji/pull-request/841 which makes this value configurable in kojid.conf
I would like to:
- Update koji on buildvm-s390x-02 to the version I just built in
rawhide/f28 that has a backport of the patch that makes this value configurable.
- Apply the following ansible patch to set the timeout only on the
buildvm-s390x-02 builder:
diff --git a/roles/koji_builder/templates/kojid.conf b/roles/koji_builder/templates/kojid.conf index 832d80c..9f10d9c 100644 --- a/roles/koji_builder/templates/kojid.conf +++ b/roles/koji_builder/templates/kojid.conf @@ -20,7 +20,12 @@ maxjobs=25 keepalive=False
rpmbuild_timeout=172800
+{% if ansible_hostname.startswith('buildvm-s390x-02') %} +; Set oz timeout higher on x390x image builder to allow it to finish. +; Install timeout(seconds) for image build +; if it's unset, use the number in /etc/oz/oz.cfg, supported since oz-0.16.0 +oz_install_timeout=14400 +{% endif %} use_createrepo_c=True
{% if host in groups['buildvm-s390x'] %}
- Run the playbook on it to update the config and restart kojid.
+1s?
kevin
infrastructure mailing list -- infrastructure@lists.fedoraproject.org To unsubscribe send an email to infrastructure-leave@lists.fedoraproject.org
infrastructure mailing list -- infrastructure@lists.fedoraproject.org To unsubscribe send an email to infrastructure-leave@lists.fedoraproject.org
infrastructure mailing list -- infrastructure@lists.fedoraproject.org To unsubscribe send an email to infrastructure-leave@lists.fedoraproject.org
On 03/16/2018 03:29 PM, Kevin Fenzi wrote:
Greetings.
image composes using oz take a long time on our s390x image builder, (due to them being in another datacenter). They take so long, they timeout.
See: https://pagure.io/koji/pull-request/837 from sinny and https://pagure.io/koji/pull-request/841 which makes this value configurable in kojid.conf
I would like to:
- Update koji on buildvm-s390x-02 to the version I just built in
rawhide/f28 that has a backport of the patch that makes this value configurable.
- Apply the following ansible patch to set the timeout only on the
buildvm-s390x-02 builder:
Thanks Kevin. We are now hitting another issue on s390x during anaconda.
https://koji.fedoraproject.org/koji/taskinfo?taskID=25761596 ``` 10:58:45,692 ERR anaconda:anaconda: startup_utils: Waiting for modules to be started timed out. 10:58:45,804 ERR anaconda:anaconda: stdout: Anaconda DBus modules failed to start on time. ```
Sinny, can you investigate this new issue?
Dusty
On Sat, Mar 17, 2018 at 8:48 PM, Dusty Mabe dusty@dustymabe.com wrote:
On 03/16/2018 03:29 PM, Kevin Fenzi wrote:
Greetings.
image composes using oz take a long time on our s390x image builder, (due to them being in another datacenter). They take so long, they timeout.
See: https://pagure.io/koji/pull-request/837 from sinny and https://pagure.io/koji/pull-request/841 which makes this value configurable in kojid.conf
I would like to:
- Update koji on buildvm-s390x-02 to the version I just built in
rawhide/f28 that has a backport of the patch that makes this value configurable.
- Apply the following ansible patch to set the timeout only on the
buildvm-s390x-02 builder:
Thanks Kevin. We are now hitting another issue on s390x during anaconda.
https://koji.fedoraproject.org/koji/taskinfo?taskID=25761596
10:58:45,692 ERR anaconda:anaconda: startup_utils: Waiting for modules to be started timed out. 10:58:45,804 ERR anaconda:anaconda: stdout: Anaconda DBus modules failed to start on time.Sinny, can you investigate this new issue?
Last two F28 S390x cloud-base compose from 20180317 [1] and 20180318[2] looks good. It might be a temporary issue, will look into it if error occurs again.
[1] https://kojipkgs.fedoraproject.org/compose/branched/Fedora-28-20180317.n.0/c... [2] https://kojipkgs.fedoraproject.org/compose/branched/Fedora-28-20180318.n.0/c...
Dusty _______________________________________________ infrastructure mailing list -- infrastructure@lists.fedoraproject.org To unsubscribe send an email to infrastructure-leave@lists.fedoraproject.org
On Sat, Mar 17, 2018 at 12:59 AM, Kevin Fenzi kevin@scrye.com wrote:
Greetings.
image composes using oz take a long time on our s390x image builder, (due to them being in another datacenter). They take so long, they timeout.
See: https://pagure.io/koji/pull-request/837 from sinny and https://pagure.io/koji/pull-request/841 which makes this value configurable in kojid.conf
Thanks a lot Kevin for this fix!
I would like to:
- Update koji on buildvm-s390x-02 to the version I just built in
rawhide/f28 that has a backport of the patch that makes this value configurable.
- Apply the following ansible patch to set the timeout only on the
buildvm-s390x-02 builder:
diff --git a/roles/koji_builder/templates/kojid.conf b/roles/koji_builder/templates/kojid.conf index 832d80c..9f10d9c 100644 --- a/roles/koji_builder/templates/kojid.conf +++ b/roles/koji_builder/templates/kojid.conf @@ -20,7 +20,12 @@ maxjobs=25 keepalive=False
rpmbuild_timeout=172800
+{% if ansible_hostname.startswith('buildvm-s390x-02') %} +; Set oz timeout higher on x390x image builder to allow it to finish. +; Install timeout(seconds) for image build +; if it's unset, use the number in /etc/oz/oz.cfg, supported since oz-0.16.0 +oz_install_timeout=14400 +{% endif %} use_createrepo_c=True
{% if host in groups['buildvm-s390x'] %}
- Run the playbook on it to update the config and restart kojid.
+1s?
kevin
infrastructure mailing list -- infrastructure@lists.fedoraproject.org To unsubscribe send an email to infrastructure-leave@lists.fedoraproject.org
On 03/18/2018 08:33 PM, Sinny Kumari wrote:
On Sat, Mar 17, 2018 at 12:59 AM, Kevin Fenzi kevin@scrye.com wrote:
Greetings.
image composes using oz take a long time on our s390x image builder, (due to them being in another datacenter). They take so long, they timeout.
See: https://pagure.io/koji/pull-request/837 from sinny and https://pagure.io/koji/pull-request/841 which makes this value configurable in kojid.conf
Thanks a lot Kevin for this fix!
No problem. :)
kevin
infrastructure@lists.fedoraproject.org