Merge ~rbalint/autopkgtest-cloud:systemd-big-amd64 into autopkgtest-cloud:master

Proposed by Balint Reczey
Status: Merged
Merged at revision: 68e47697a6dc2ffd8d7e7cce9d04728714799954
Proposed branch: ~rbalint/autopkgtest-cloud:systemd-big-amd64
Merge into: autopkgtest-cloud:master
Diff against target: 12 lines (+1/-1)
1 file modified
worker-config-production/worker.conf (+1/-1)
Reviewer Review Type Date Requested Status
Iain Lane Approve
Review via email: mp+397740@code.launchpad.net
To post a comment you must log in.
Revision history for this message
Christian Ehrhardt  (paelzer) wrote :

Systemd tests are always a bit problematic, and getting them to be more stable would help a lot.
While marking them big will make them wait longer for free resources and block more I think it is worth also for reducing the time of a single run. I'm crossing fingers it will help with the current issues as you hope.

To me it is always confusing which files control what, but for a full x86 change don't you need to also modify these?
 worker-bos01-arm64.conf
 worker-bos02-arm64.conf
 worker-canonistack.conf

Revision history for this message
Balint Reczey (rbalint) wrote :

*-arm64.conf files are for ARM64, but it is easy to read them as -amd64 at first glance. :-)

Revision history for this message
Balint Reczey (rbalint) wrote :

 worker-canonistack.conf lists only releases = trusty xenial. I'm not sure it a change there would affect hirsute tests.

Revision history for this message
Iain Lane (laney) wrote :

That's right, that config file is only an example for if you wanted to run the whole stack in Canonistack.

My position is that if this gets us consistently passing tests - avoiding the burden of people having to retry all the time and wait for ages - then we should do it.

I've been running tests in the production infrastructure to see how much this will help. I have run 3 complete runs on "m1.large" instances. One of them succeeded, the other two failed. A 4th is going now, and I'll post its log once that is done (might be tomorrow depending on how it goes).

I'm now running the same in Canonistack, on m1.large. My hope is that this fails in a similar way, which should be more helpful for debugging, as apparently these failures are not reproducible for people locally.

Question while we're waiting for that: what is it about amd64 that makes this necessary when it's not needed for other arches? Or to ask in a different way, if we're going to do it in one place, should we do it everywhere for sanity?

review: Needs Information
Revision history for this message
Balint Reczey (rbalint) wrote :

This is a good question. There is no big differences between architecures regarding tests (except for armhf due to not running serveral tests in containers).

Christian observed many timeout that can come from the severe oversubscription of the test infra nodes or kernel/virtualization issues but I can't tell.

Revision history for this message
Iain Lane (laney) wrote :

Let's keep updates on LP: #1915126, posting there now.

Revision history for this message
Dan Streetman (ddstreet) wrote :

pretty please include systemd-upstream in big tests too?

> what is it about amd64 that makes this necessary when it's not needed for other arches

the integration tests ('upstream' tests) only run under qemu on amd64 and ppc64el, currently, so the that test suite will go much faster on other archs (since it skips all the under-qemu testing). That'll change once we can move to focal and expand use of qemu to all archs, which is waiting on lp: #1895132

Revision history for this message
Balint Reczey (rbalint) wrote :

Please make systemd run on big VMs temporarily and I'll split the upstream test to two halves which can be SRUd and then systemd won't time out on small VMs either.

Revision history for this message
Iain Lane (laney) :
review: Approve

Preview Diff

[H/L] Next/Prev Comment, [J/K] Next/Prev File, [N/P] Next/Prev Hunk
diff --git a/worker-config-production/worker.conf b/worker-config-production/worker.conf
index 9d7c435..fdba3df 100644
--- a/worker-config-production/worker.conf
+++ b/worker-config-production/worker.conf
@@ -16,7 +16,7 @@ releases = precise trusty xenial bionic focal groovy hirsute
16architectures = i386 amd6416architectures = i386 amd64
17setup_command = $AUTOPKGTEST_CLOUD_DIR/worker-config-production/setup-canonical.sh17setup_command = $AUTOPKGTEST_CLOUD_DIR/worker-config-production/setup-canonical.sh
18setup_command2 = $CHECKOUTDIR/setup-commands/setup-testbed18setup_command2 = $CHECKOUTDIR/setup-commands/setup-testbed
19big_packages = binutils chromium-browser glibc libreoffice linux linux-* tdb firefox akonadi julia meson python-cryptography python-dmsh pyresample libtext-bidi-perl python-hypothesis nova rocs camitk kineticstools r-bioc-variantannotation r-cran-igraph botch mathicgb ganeti lttng-modules python2.7 python3.6 python3.7 python3.8 python-ruffus openjdk-8 openjdk-lts openjdk-13 openjdk-14 nodejs cl-ironclad python-scipy scipy cross-toolchain-base-mipsen mkl-dnn samtools satpy python-cffi yt skimage scikit-learn silx statsmodels puma multipath-tools tgt dask assimp mariadb-10.3 mariadb-10.5 deal.ii libsis-jhdf5-java octave-parallel r-bioc-beachmat19big_packages = binutils chromium-browser glibc libreoffice linux linux-* tdb firefox akonadi julia meson python-cryptography python-dmsh pyresample libtext-bidi-perl python-hypothesis nova rocs camitk kineticstools r-bioc-variantannotation r-cran-igraph botch mathicgb ganeti lttng-modules python2.7 python3.6 python3.7 python3.8 python-ruffus openjdk-8 openjdk-lts openjdk-13 openjdk-14 nodejs cl-ironclad python-scipy scipy cross-toolchain-base-mipsen mkl-dnn samtools satpy python-cffi yt skimage scikit-learn silx statsmodels puma multipath-tools tgt dask assimp mariadb-10.3 mariadb-10.5 deal.ii libsis-jhdf5-java octave-parallel r-bioc-beachmat systemd
20long_tests = gutenprint gmp-ecm open-iscsi diffoscope snapcraft cross-toolchain-base mercurial20long_tests = gutenprint gmp-ecm open-iscsi diffoscope snapcraft cross-toolchain-base mercurial
21blacklist = bionic/i386/upstart bionic/amd64/upstart bionic/i386/libnih bionic/amd64/libnih bionic/i386/r-cran-vioplot bionic/amd64/r-cran-vioplot bionic/amd64/stress-ng bionic/i386/stress-ng bionic/i386/openjdk-lts xenial/amd64/virtualbox-ext-pack bionic/amd64/virtualbox-ext-pack xenial/i386/virtualbox-ext-pack bionic/i386/virtualbox-ext-pack hirsute/i386/rust-onig hirsute/amd64/rust-onig21blacklist = bionic/i386/upstart bionic/amd64/upstart bionic/i386/libnih bionic/amd64/libnih bionic/i386/r-cran-vioplot bionic/amd64/r-cran-vioplot bionic/amd64/stress-ng bionic/i386/stress-ng bionic/i386/openjdk-lts xenial/amd64/virtualbox-ext-pack bionic/amd64/virtualbox-ext-pack xenial/i386/virtualbox-ext-pack bionic/i386/virtualbox-ext-pack hirsute/i386/rust-onig hirsute/amd64/rust-onig
2222

Subscribers

People subscribed via source and target branches