Ken Dreyer [Tue, 26 May 2015 20:17:04 +0000 (14:17 -0600)]
remove fedora20 Jenkins slave
Remove the Fedora 20 build slave from the job configs.
We're having trouble resurrecting
gitbuilder-cdep-rpm-cloud-fedora20-amd64-basic.cloud.sepia.ceph.com in
DreamCompute, and Fedora 20 itself is going EOL in a month or so.
This is the only Fedora builder we have, and this VM does build packages
for http://ceph.com/rpm-hammer/fc20/, so it would be nice to get fresh
Fedora build slaves set up eventually. But in the meantime, since this
one is giving us trouble, just remove it.
Ken Dreyer [Tue, 5 May 2015 15:40:27 +0000 (09:40 -0600)]
ceph-release-rpm: update EL6 references to "6.5"
This was an older job that didn't get updated often, so when we added
the "6.5" builders, this job was neglected.
Update all the "6.3" references to "6.5" so we can drop the old "6.3"
buildslaves. EL 6.5 has been EOL for a while now and is no longer
receiving security updates.
Ken Dreyer [Mon, 4 May 2015 23:46:36 +0000 (17:46 -0600)]
unquote archive settings
Prior to this commit the archive settings for ceph-deploy and
radosgw-agent were quoted 'false' or 'true' instead of simply the
boolean types that YAML expects.
jenkins-jobs interpreted these strings to be "true", so we ended up with
the following XML:
<latestOnly>true</latestOnly>
This meant that Jenkins deleted all the older archived ceph-deploy
packages each time a new version was built. When 1.5.22.1 was built,
the "older" 1.5.23 package was deleted.
Unquote the boolean values so JJB will properly interpret them.
Ken Dreyer [Tue, 14 Apr 2015 17:22:29 +0000 (11:22 -0600)]
ceph-setup: remove do_autogen.sh
Remove do_autogen.sh from the ceph-setup task. Run ./autogen.sh and
./configure instead.
I've copied this long ./configure argument list from a prior run of
./do_autogen.sh so that we're using the same arguments. We can adjust it
as needed.
One of the reasons for making this change is that ./do_autogen.sh takes
an argument ("-L") depending on whether we're building a Hammer or
Firefly release. The goal of this change is to use a more stable
interface for doing the builds.
As background, from Sage on IRC yesterday:
sage: i'm surprised jenkins is using do_autogen.sh at all actually?
ktdreyer: what should it use?
sage: ./autogen.sh and then pass the args explicitly to ./configure
that it wants?
sage: it doesn't really matter i guess, but do_autogen.sh piles a
bunch of random stuff in there
ktdreyer: I was just looking at autobuild-ceph.git's build-ceph.sh
ktdreyer: it does what you're suggesting
ktdreyer: cool
sage: do_autogen.sh in my mind is for developers. if it makes sense
for relase that's ok, but that's not how i've treated it
Ken Dreyer [Mon, 13 Apr 2015 23:06:23 +0000 (17:06 -0600)]
ceph-build-pull-requests: quiet output
Running the pull requests script with -x is very noisy. The script
already prints a lot of information with "echo", so just silence the
debugging outupt.
Ken Dreyer [Mon, 30 Mar 2015 23:36:03 +0000 (17:36 -0600)]
build_rpms.sh: terminate on unpackaged files
As Ceph changes and new files get installed by default, we need to
ensure that the commits which create these files also update the
packaging files lists as well.
One way to do this is to tell Gitbuilder to stop skipping unpackaged
files in the RPMs. This will cause Gitbuilder builds to indicate a
failure when new files get added without packaging updates.
Ken Dreyer [Mon, 1 Dec 2014 18:15:07 +0000 (11:15 -0700)]
doc: add notes about "documentation" jobs
Add a bit of documentation regarding "documentation" jobs in Jenkins.
This records our lessons learned during Andrew's work to create the
"teuthology-docs-build" job configuration and get that particular one
running.
Ken Dreyer [Mon, 17 Nov 2014 21:42:36 +0000 (14:42 -0700)]
ceph-deploy-pull-requests: add Travis to admins list
Travis is a co-maintainer of ceph-deploy; add him to the list of
administrators for the GitHub Pull Requests plugin. Jenkins will trust
any pull requests from Travis and build them automatically.
(I don't know why Jenkins wasn't already trusting Travis since he's a
member of the Ceph organization.)
Ken Dreyer [Fri, 14 Nov 2014 15:13:40 +0000 (08:13 -0700)]
jjb: increase precision of job searches
The "for `ls .` ... -d subdirectory" logic was examining every entry at
the top-level of the workspace, both files and directories. The files
entries were useless since they would never contain subdirectories.
Use "find" instead to only locate directories at the top level.
The purpose of this change is to make the verbose output (with set -x)
easier to read, and make it easier to spot errors in the future.
Ken Dreyer [Fri, 14 Nov 2014 14:27:53 +0000 (07:27 -0700)]
jjb: be verbose and fail early on errors
Prior to this commit, if the "job-builder update" command failed,
Jenkins would still consider the overall job to have succeeded, because
the jjb/config/config script was exiting with a zero exit code. (This
was because the for/ls loop were the final commands to run, and they
were exiting with zero.)
When I modified the Jenkins UID as a part of the overall Jenkins
authentication changes yesterday, the "job-builder update" command no
longer worked, because the password in jenkins_builder.ini was no longer
valid. However, Jenkins still considered the jjb job to be passing, as
explained above.
Tell bash to be verbose about the commands it runs ("set -x") and fail
early if there are any problems ("set -e").
With this change, if there is a problem with "job-builder update" or any
other command in the script, the job will fail right away and Jenkins
will bring it to our attention sooner.
projects / ceph-build.git / log