Age | Commit message (Collapse) | Author | Files | Lines |
|
MCP does not have a self-test healtcheck suite, similar to old
Fuel, so we dropped the "-H" deploy argument (which used to disable
healtcheck autorun).
Adjust deploy command to reflect this installer change.
While at it, fix leftover "pxebr" default installer parameter,
which we missed in the previous commit where we changed the
default only in releng deploy scripts.
Change-Id: I55e1f5cad17687bdf304fbf04d67366dd32ed42a
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Fuel and implicitly Armband switched to a newer format for parsing
the "-B" deploy arg, used to pass bridge(s) information.
For now, we will default to ",,,", which is equivalent to not
setting any default value for any of the 4 possible bridges.
NOTE: Armband overrides this on a per-pod basis, since bridge
information is POD specific in our lab.
Change-Id: I3635b0eb6f8233cf99022a97d5642f2e49197e3b
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Old Fuel deploy scripts required the virtual POD name to be
virtual_kvm, which needed to be translated at CI level after
source enviroment specific configuration.
New Fuel deploy scripts no longer need this translation, so
drop it.
Change-Id: I18164e77895bbb03a1ca7798e3c0ba78c34d1c65
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Using sudo triggers a chain of dependencies in our infrastructure,
ranging from issues cleaning up the Jenkins workspace to deploy-time
permission issues.
As it turns out, cleaning all looses ends after that change is not a
trivial task, and the remaining work outweights the benefits of using
sudo in the first place.
The original motivation for using sudo was a superposition of:
- hardcoded check for running as root in Fuel's ci/deploy.sh;
- the ability to install packages on the hosts via apt-get/yum;
- the ability to spawn VMs using virt-manager;
All of the above can be mitigated by imposing a series of restrictions
for the Jenkins slave / jump server workstation and the jenkins user:
- check for sudo rights instead of running as root user;
- explicitly use sudo for package installation and/or system-level
changes;
- add jenkins user to "kvm" and "libvirtd" groups;
So, revert using sudo when calling the deploy script, and limit sudo
actions to package installation in Armband's deploy script.
To compensate running as a regular user, a series of small changes is
required in Armband project itself, together with the appropiate
Jenkins slave user configuration on the machines running this script.
This reverts commit 67d8bac7b2a97531ad465f275747b3dc455d4fe4.
Change-Id: I88df6db1a43b93b7314ceca53d353f9000f15153
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Previous commit adding sudo for ci/deploy.sh call breaks Armband
deploys, since ci/deploy.sh relies on $WORKSPACE env var being set.
Change-Id: I948c8e3c00b430e0c186c0a727e1b7e027d7a089
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
New MCP deploy scripts have a hard requirement on running as root.
Bring back sudo for the `ci/deploy.sh` invocation, aligning with
Fuel@OPNFV.
NOTE: This can be later avoided/reverted if we can guarantee the
"jenkins" user has sudo rights, even without a tty present.
Change-Id: Ibec7b14e9174a7f482609890e5e68912931a8d05
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Sync with changes for Fuel@OPNFV targeting master branch (MCP):
- disable ISO build for master branch;
- turn off upload action & checkout;
- add retrieving OS credentials for MCP;
- disable ISO download for master branch;
Change-Id: I11eacfad7ee1f738119e17c2e63c3d92707f8e1f
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Arm-pod3-2 is a second jenkins slave for arm-pod3.
This slave is arm while the old jenkins slave is x86.
Fatih requested for this slave to be renamed to arm-pod4
instead of arm-pod3-2, even if it's a slave for an existing pod
JIRA: ARMBAND-252
Change-Id: I5645edd7d774f763e5e72115851c7061bfafd344
Signed-off-by: Cristina Pauna <cristina.pauna@enea.com>
|
|
arm-pod3-2 is an aarch64 jenkins slave for the same POD as the existing
x86 jenkins slave arm-pod3. This commit adds jobs so we can deploy the
POD and run functest from the aarch64 jenkins slave.
These jobs are non-ci
JIRA: ARMBAND-241
Change-Id: Ie3b808a641e44649c02f3b75c83d79119217fdfe
Signed-off-by: Cristina Pauna <cristina.pauna@enea.com>
|
|
Fuel and Armband deploy scripts trap internal errors and call exit
in the trap handler, so the releng wrapper exists before collecting
all deploy logs when shell option "errexit" is set.
JIRA: ARMBAND-217
Change-Id: Icb1e3277c3e137abb3225a2b9a0b4269eaf08fc6
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
GIT_BRANCH parameter is different; master or stable/danube in daily
jobs, but 'refs/changes/57/27657/6' in verify jobs. This breaks some
job builders triggered by the both types of jobs. We have GERRIT_BRANCH
parameter for verify jobs to identify stream and expected branch into
the patch will be merged after it got +2 and submitted.
To avoid further confusion and to have common job builders for daily
and verify jobs, this patch introduce BRANCH parameter.
GERRIT_BRANCH is now deprecated.
Change-Id: Ibcd42c1cd8a0be0f330878b21d3011f1ec97043b
Signed-off-by: Ryota MIBU <r-mibu@cq.jp.nec.com>
|
|
This enables the mechanism added in [1].
Armband just added support for caching deploy artifacts, like
bootstraps and target images, which take a lot of time at each deploy
to be built, considering it requires a cross-debootstrap via
qemu-user-static and binfmt.
For easier setup, the deploy cache depends on the local ISO storage
mechanism (/iso_mount/opnfv_ci), and reuses the same storage pool.
To enable deploy caching, just create the deploy-cache directory at:
/iso_mount/opnfv_ci/<branch>/deploy-cache
JIRA: ARMBAND-172
[1] https://gerrit.opnfv.org/gerrit/#/c/25191
Change-Id: I161205a87f13cd78e79b96bfc37a5f68aba093cc
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Pod virtual1 was added in the ENEA-ARM lab
This commit adds the jobs for it
JIRA: ARMBAND-162
Change-Id: Icaa798691f5ff60a58ae083eb1c366170c464b92
Signed-off-by: cristinapauna <cristina.pauna@enea.com>
|
|
This depends on:
- jenkins-enea ssh access to OPNFV gerrit;
- master and stable/colorado securedlab definitions for arm-pod* [1];
[1] https://gerrit.opnfv.org/gerrit/#/c/23159/
JIRA: ARMBAND-149
Change-Id: Ic9a2e232027636694ffd441fe5203cb06975f137
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Since we do not actually parse our build/deploy logs for color chars,
set the terminal to "vt220".
This will silence quite a few of colorization special identifiers,
but NOT all of them (e.g. `tput sgr0` still prints a sequence).
For Fuel@OPNFV build process, we also need to workaround docker
terminal settings not being passed down when using a tty (see [1]),
which will be fixed in Docker 1.13+. This fix is handled in [2].
[1] https://github.com/docker/docker/issues/9299
[2] https://gerrit.opnfv.org/gerrit/#/c/22883/
Change-Id: I5f0c6eae33349ef81b67dd17e6df4e0af6274fab
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Closes-bug: ARMBAND-72
Change-Id: I6c5a3e6ce4c320bf42f02b290a0137f5b892ad7f
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Previous change adding lab-config/.../local_env support introduced
an issue, in the form of `test -e file:///path/to/local_env`,
which is not supported.
Change-Id: I1566b41765329f18523cfc506c9112b6694a8317
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
After previous change introducing the usage of node labels for
Armband CI deploy jobs, we discovered that the old params were
not applied to labeled pods.
To complicate things, we had different DEFAULT_BRIDGE params
for each CI pod, which contradicts the Jenkins templating
paradigm.
Since we already have a mechanism in place for overriding
env vars based on the POD we are running against, move both
<DEPLOY_TIMEOUT> and <DEFAULT_BRIDGE> params to our local_env,
which is sourced by the deploy script and can apply POD specific
configuration.
Add missing LAB_CONFIG_URL to newly added baremetal and virtual
defaults for Armband.
While at it, also replace "brahmaputra" comments with "colorado".
v2 -> v3:
Source local_env before using DEFAULT_BRIDGE in <armband-deploy.sh>.
Previously, we (Armband) sourced this file in our <ci/deploy.sh> [1],
which will be dropped after this change is merged.
[1] https://git.opnfv.org/cgit/armband/tree/ci/deploy.sh#n24
Change-Id: Ibafea25b3df8c3118ce05e3f40934c8e8b9c8dcd
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Change-Id: I317321bbaf1e74419b9508c6da3039b95d9d399d
Signed-off-by: Josep Puigdemont <josep.puigdemont@enea.com>
|
|
The download script will now use gsutils if they are available, or fall
back to curl as before.
In general, use curl -f, which will exit with an error code if the
server gave an error, instead of writing a file with the HTML error
page.
Change-Id: Ib24e7803ec38b20a8879cd1509a0cb7b512fee61
Signed-off-by: Josep Puigdemont <josep.puigdemont@enea.com>
|
|
Change-Id: I76b0aadb735eefc13bfe9364b7426d0a28aa9b96
Signed-off-by: Josep Puigdemont <josep.puigdemont@enea.com>
|
|
Change-Id: I5b261885585fd9e902b3a7513a6d07881e849976
Signed-off-by: Josep Puigdemont <josep.puigdemont@enea.com>
|
|
Change-Id: I6918dc1189ab18fa79812ffa7c5e67867a945eba
Signed-off-by: Josep Puigdemont <josep.puigdemont@enea.com>
|