Dan Mick [Tue, 3 Dec 2024 22:05:24 +0000 (14:05 -0800)]
scripts/get-tarballs.sh: pull to prerelease area
I've got a bunch of doc work to do to the release process;
this is one of them. Maybe we should also add this, and
its copy, to the ceph-cm-ansible role for d.c.c
Dan Mick [Tue, 3 Dec 2024 19:13:35 +0000 (11:13 -0800)]
Keep track of extra script get-tarballs.sh
This is used by sync-pull to fetch the tarballs when
constructing a release, but needs modification for the
prerelease/ directory structure. It's invoked by ssh
to download.ceph.com to run from /home/signer/bin, because
it needs to copy large files from chacra to download.ceph.com
and doing that on download.ceph.com is much more efficient.
Nizamudeen A [Tue, 12 Nov 2024 05:30:06 +0000 (11:00 +0530)]
sync-images: sync all images including their different aarch
The sync is not syncing the different architectures supported by the
images atleast for grafana. And hence only the amd64 gets mirrored to
the quay repo and causes issues to arm64 builds. I saw that skopeo
supports an `--all` opton that copies every aarch.
ceph.git's build.sh does not handle authentication in its latest
version, but assumes it's been done. Add the authentication
step (podman login) here, in setup_rpm, with the side benefit
that it'll fail faster if the credentials are incorrect.
... as they are ignored in scripts/build_utils.sh since commit 3517941ef2aa ("script/build_utils.sh: don't fail if any mirror is not
accessible"). This is prompted by an issue with libboost repo, which
started returning 403 Forbidden errors.
While at it, skip fetching language and translation packs.
ceph-volume: Rename ceph-volume-linter to ceph-volume-unit-tests
This:
- Update the job name and display name in ceph-volume-pr.yml to reflect the change.
- Changes the trigger phrase and GitHub status context for unit tests.
Dan Mick [Thu, 19 Sep 2024 00:55:56 +0000 (17:55 -0700)]
ceph-dev-new-*: change for in-tree container builds
- invoke ${SRC}/container/build.sh to build the container
- do not check out ceph-container.git
- add test to loop for "is repo ready" to test for the jenkins job
url ${BUILD_URL} as well as ready, so that we're not racing against
preexisting repos for this branch/os/flavor
Dan Mick [Mon, 19 Aug 2024 03:02:14 +0000 (20:02 -0700)]
sepia-fog-images: allow ansible to find collections
The mount module is now ansible.posix.mount, and ansible-playbook
can't seem to find it when executed from the parent directory.
Set ANSIBLE_CONFIG, which sets the collection search path.
Due the the EOL of CentOS Stream 8, all the Jenkins workers were
migrated to CentOS Stream 9. Therefore, we need to make these jobs pick
these workers.
Dan Mick [Tue, 2 Jul 2024 22:18:37 +0000 (15:18 -0700)]
prune-quay.py: avoid bug deleting -crimson tags
<sha1>-crimson tags stand sort of alone. Don't delete them when
looking for sha1 tags, unless their x86 counterpart (<sha1>) is
going to be deleted as well.
Dan Mick [Tue, 2 Jul 2024 06:49:17 +0000 (23:49 -0700)]
prune-quay.py: remove short_sha1_cache
Not only is it no longer necessary now that we're caching the
entire shaman result, it actually can lead to missing deletions
if the same sha1 validly exists for different builds (like for
centos8 vs centos9, or crimson vs default, etc.)
Dan Mick [Tue, 2 Jul 2024 06:28:58 +0000 (23:28 -0700)]
prune-quay.py: rely on delete-all-with-same-digest
With the new "delete all tags with the same digest", there's no need to
keep the short_sha1 around to match against sha1 tags; they'll be
caught earlier by the 'same-digest' test.
Dan Mick [Fri, 28 Jun 2024 21:58:50 +0000 (14:58 -0700)]
prune-quay.py: stop using image_id, optimize 'same image' search
image_id has apparently gone away; take the opportunity to be
a bit more efficient: keep a map of digests to tagnames, and when
deleting a full tag, also delete any tags that have the same
digest. Only verbosely report if any new tags were marked.
This commit makes the job sync nginx images from docker.io
to quay.io
Also, two changes:
- This makes sure the latest stable image of skopeo is pulled from quay.io
- Use the filter 'by-semver' filter for grafana too.
Nizamudeen A [Fri, 14 Jun 2024 04:02:58 +0000 (09:32 +0530)]
scripts/dashboard: kill existing xvfb process
To fix the below error on starting cypress
```
[STARTED] Task without title.
[FAILED] Xvfb exited with a non zero exit code.
[FAILED]
[FAILED] There was a problem spawning Xvfb.
[FAILED]
[FAILED] This is likely a problem with your system, permissions, or installation of Xvfb.
[FAILED]
[FAILED] ----------
[FAILED]
[FAILED] Error: _XSERVTransSocketUNIXCreateListener: ...SocketCreateListener() failed
[FAILED] _XSERVTransMakeAllCOTSServerListeners: server already running
[FAILED] (EE)
[FAILED] Fatal server error:
[FAILED] (EE) Cannot establish any listening sockets - Make sure an X server isn't already running(EE)
[FAILED]
[FAILED] ----------
[FAILED]
[FAILED] Platform: linux-x64 (Ubuntu - 22.04)
[FAILED] Cypress Version: 12.17.4
Xvfb exited with a non zero exit code.
There was a problem spawning Xvfb.
This is likely a problem with your system, permissions, or installation of Xvfb.
```
Ilya Dryomov [Fri, 31 May 2024 21:51:49 +0000 (23:51 +0200)]
ceph-windows-pull-requests: constrain to windows label
Currently, the intersection of focal && libvirt labels returns some
smithi (nproc == 8) and mira (nproc == 4) nodes, which are inadequate
for the two VMs that the Windows job spawns. It turns we already have
a windows label, which, after dropping it from mira015, seems to be on
point (i.e. would filter those low CPU nodes out). Let's use it!
Add parameters to set variables that the job scripts will look for
to choose which flavors to build. This will override the defaults in
ceph-container.git; the defaults are set to the same as
ceph-container.git at this moment, but are probably soon to change to
centos,9.
projects / ceph-build.git / log