Age | Commit message (Collapse) | Author | Files | Lines |
|
Build working, automated deploy working (simple test cases only).
All plugins disabled, so is the generation of the local ISO cache.
Both need to be re-enabled after merge of this uplift.
All patches wiped, pertinent ones need to be re-enabled after the
merge of this uplift.
Change-Id: I978a3d4ed25a45a4b88a6e35693baac597b162c2
Signed-off-by: Stefan K. Berg <stefan.k.berg@ericsson.com>
|
|
Tests shows that networking-odl v2 introduce more problems when it comes
to propagate SG rules on ovs. Disable v2 for all scenarios.
See: https://bugs.opendaylight.org/show_bug.cgi?id=7021
JIRA: FUEL-216
Change-Id: I5440ba7ff81b2063491259d91f17cb5578a8c48c
Signed-off-by: Michal Skalski <mskalski@mirantis.com>
|
|
|
|
|
|
* use raw devices instead of qcow2
* disable i/o caching, it reduces data copies and bus traffic
* set native aio instead of thread based
JIRA: FUEL-201
Change-Id: I8ac728cbb5be5948080fab8ae4d3d311e0a5c09c
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
Split out public_vip_ping task to avoid race condition when ping_checker
resource might be started up on controllers before a primary one.
JIRA: FUEL-217
Change-Id: I3c11e8403d4e1c53a7f0eac07c71f59584882332
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
JIRA: FUEL-215
Change-Id: Ie922ff794e686e1824483a91a9571252adc2b1e2
Signed-off-by: Fedor Zhadaev <fzhadaev@mirantis.com>
|
|
Avoid public vip failover at the end of post-deployment
by moving ping_checker resource into deployment stage.
Also pingd constraint has been changed to be based on score.
If all the nodes failed to ping the default gw, the resource
ain't stopped and remains on the last node where it was.
JIRA: FUEL-212
JIRA: FUEL-217
Change-Id: I09329474cf13958499ac99bfbdf9c9055c9bcb55
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
|
|
JIRA: FUEL-212
Change-Id: Ie8c50d515bffbb39deb44c32da42079b4181edd3
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
|
|
Repoint to the "Update flavors mem_page_size"
from stable/9.0 branch of plugin.
Change-Id: I91d6a770faa9ee2d078b7f1ccccb7d61525ad923
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
To properly handle source routing passing by dhcpd
inside instance get latest cirros image.
JIRA: FUEL-202
Change-Id: Icd44ebbdd7b4aa0cb55c080e614616a7b0198c0f
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
JIRA: FUEL-212
Change-Id: I1c5b8fe123406f1a96c5432d91a9ee5806c7cf62
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
|
|
|
|
Fuel 9.1 was released, it is a first time when new version of Fuel is
distributed through the same repositories as previous one, point to
specific snaphost of mos ubuntu repo to prevent mixing packages.
Change-Id: I64bb52657ce476b0c1fdc8e858231ad751ba12b1
Signed-off-by: Michal Skalski <mskalski@mirantis.com>
|
|
|
|
Fixes https://jira.opnfv.org/browse/FUEL-211
Change-Id: Iebb477b2a4ff370198d05edf595b5059ba47f51a
Signed-off-by: Peter Barabas <peter.barabas@ericsson.com>
|
|
|
|
|
|
|
|
|
|
After Armband rework landed of top of `build/f_repos` mechanism [1],
the following minor UX improvements can also be applied to Fuel's
f_repos:
- set SHELL to "/bin/sh" (we don't use any bash-isms);
- improve "From SHA..." removal for exported patches;
- force checkout of root tag commit during clean;
- silence progress during git clone (cleaner logs);
- support git older than 1.8.4 in `make clean` (fix Armband deploy);
FIXME:
Pass TERM as Docker env var until [2] is fixed in Docker 1.13.
v4 -> v5:
* Moved `export TERM` to Releng, where it actually belongs [3];
* Fixed Armband deploys fail [4] by adjusting `make clean`;
[1] https://gerrit.opnfv.org/gerrit/#/c/22791/
[2] https://github.com/docker/docker/issues/9299
[3] https://gerrit.opnfv.org/gerrit/#/c/22933/
[4] https://build.opnfv.org/ci/view/armband/job/\
fuel-deploy-armband-baremetal-daily-master/57/consoleText
Fixes: FUEL-200
Change-Id: I80e3074f8659769e21f5b56f07c34c7a5de727bc
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
ODL Boron will be used in all scenarios except bgpvpn beacuse vpnservice
was removed in this release, we still will use Beryllium for this
scenario.
This change only replace Beryllium with Boron, it does not use any new
features like odl-netvirt-openstack.
Change-Id: I95264cc97f5b9e922ab3fb7dc024724d9fb83e9c
Signed-off-by: Michal Skalski <mskalski@mirantis.com>
|
|
Some mirrors are misbehaving, for instance due to flood protection,
which makes them unreliable. This change re-introduces the possibility
to blacklist them locally.
Change-Id: I7840ac0a622f2997395e54911685f9eea393dee4
Signed-off-by: Stefan K. Berg <stefan.k.berg@ericsson.com>
|
|
Change-Id: I3ee9e5afa416af9e6a90b77f03ed9258c6bd4b1b
Signed-off-by: Guo Ruijing <ruijing.guo@intel.com>
|
|
This is the initial VM description fit for spawning Fuel Master
VMs in Enea's Pharos lab (Armband lab) [1, 2].
[1] http://www.enea.com/pharos-lab
[2] https://wiki.opnfv.org/display/pharos/Enea-pharos-lab
Change-Id: Ia89f9728fff0ab1c279f6200fd558d9643cf4ea5
Signed-off-by: Josep Puigdemont <josep.puigdemont@enea.com>
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Previous change [1] tried to move the responsibility of removing
storage files (raw, iso) from our python scripts to libvirt.
Although this works as expected for non-root invocations, like
Armband deploys, it fails to remove raw files in Fuel@OPNFV deploys.
Work around this by keeping both mechanisms (libvirt managed +
python explicit handling).
Related-bug: ARMBAND-128
[1] https://gerrit.opnfv.org/gerrit/#/c/22447
Change-Id: Ib846cd6071705b71d98824385835d724c729fd78
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
|
|
ArmbandFuel@OPNFV uses Fuel@OPNFV git repo as a submodule, so
its .git dir is located outside work tree root (F_GIT_ROOT).
Allow this setup by parsing `git rev-parse --git-dir` intead of
harcoding git dir path.
Fixes: FUEL-200
Change-Id: Ia7f3b23e9333356fe05b9ed0565218f2c7ec8eba
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
isoinfo -i lists the following information for Fuel@OPNFV ISO:
"Data preparer id: 86aafaf5454a846c417848bb94f264c4420160f3"
where the SHA hash is Fuel git repo HEAD SHA.
For Armband's build system, using only the Fuel commit ID is not
enough to fully describe the state of the source code, as patches
are also applied to other fuel modules (as git submodules).
Instead, a pointer to a valid Armband commit ID should be used.
However, Armband overrides OPNFV_GIT_SHA to the Armband git repo
commit hash, so the ISO metadata should also reflect this.
While at it, allow product name to be overriden, to signal the
commit ID should be looked up inside the Armband repository.
Change-Id: I33ad490f1afe28c1d439dda40e39cee1955e0ac2
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Change-Id: If0f5b268848439aa9d2abee099bbd985abd53d5c
Signed-off-by: Stefan K. Berg <stefan.k.berg@ericsson.com>
|
|
|
|
This change reworks support for cloning (fetching) and patching all
Fuel components (fuel-library ... fuel-ui, see [1] for full list),
both outside the ISO build process (for development purposes, like
testing OPNFV patches apply cleanly to each component), and during
the ISO build.
The implementation relies heavily on git submodules and patches,
so backporting pending upstream gerrit changes, as well as adding new
out-of-upstream-tree patches, should be trivial.
Also, since without tracking remotes the repos rarely change, leverage
the in-place OPNFV build caching mechanism to drastically reduce the
number of git clones during each build.
The mechanism is detailed more in-depth in the JIRA ticket [3], and
it's merely a rehash of the submodule handling in Armband, implemented
initially by Stanislaw Kardach <stanislaw.kardach@cavium.com>.
Pros (+ new features):
- OPNFV patches for Fuel components can be applied (imported) or
developed (exported) outside of the ISO build process;
- git repo caching reduces the number of external pulls;
- reuse the same fuel-* componets in ALL targets
e.g. fuel-mirror used to employ 2 slightly different git trees;
- one step closer to Fuel/Armband merge (lots of steps left though);
Cons:
- adds a lot complexity to solve a non-problem (handling the repos
is fine in the current form, for the current goals);
However, patching <fuel-astute> or <fuel-agent> seems to be
quite complicated with the current mechanism, as env vars
override our locally patched repos for these components).
The proposed change should eliminate this issue completely.
CHANGE:
Aligned divergent fuel-mirror in ISO vs local repo build by
applying the 302 redirect fix ("Fixed handling http redirects") [2]
on top of 9.0.1 tag in fuel-mirror repo, and using that for both.
This obsoleted the followking mk file:
- build/f_isoroot/f_repobuild/config.mk
CHANGE:
Removed unused make target `get-fuel-repo` from build/config.mk.
FIXME:
Remote tracking is currently deactivated for all branches.
Change this when remote HEAD should be tracked instead,
e.g. during Newton rebase.
[1] https://github.com/openstack/fuel-main/blob/stable/mitaka/repos.mk#L32
[2] https://review.openstack.org/#/c/334882/
[3] https://jira.opnfv.org/browse/FUEL-200
TODO: Extend build/f_repos/README.md with info about:
- branch changes;
- tag bumps;
- patching now supports subdirs;
v5 -> v6:
* Added support for nesting patches in subdirs, to be used for feature
groups and related changes across submodules.
E.g. Upcoming multiarch EFI support could go in:
- .../f_repos/patch/fuel-agent/multiarch-efi/...;
- .../f_repos/patch/fuel-astute/multiarch-efi/...;
While Armband patches could go in:
- .../f_repos/patch/fuel-agent/armband/...;
- .../f_repos/patch/fuel-astute/armband/...;
etc.
v6 -> v13:
* fix wrong tag used for VERSION dump;
* fix wrong tag used for patches-export;
* move OPNFV_TAG to main config.mk and use it in FUEL_*_COMMIT;
* only run `patches-import` when patches changed / not yet applied;
* only run `clean` for stale `.cacheid` (avoid re-applying patches);
* allow git submodule path/name to be different (FIXED);
* put-cache should only run when artifact is not already cached;
* `git am`: use 3-way merge so already applied patches are gently
skipped (useful for upstreaming patches from Armband);
* Properly support nested patch dirs;
* Colorize the output a little;
JIRA: FUEL-200
Change-Id: I9dbdd9b3022896d4497d21be8dacc859730db489
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
The previous change adding support for OPNFV build tag reuse
added the get/put-cache functionality to `build/docker` just
for detecting env var changes between tag creation and current build.
Obviously, when building outside CI, it is up to the end user to
force a docker rebuild when env vars change (this was the behavior
before the tag reuse support too).
An alternative approach would be removing get/put-cache completely,
if we consider the env vars (http proxy related) never change for
CI cached builds (seems to be true, worst case we wait 1 day for
the tag to expire when it happens).
This change just fixes the original intended behavior in tag reuse.
JIRA: FUEL-204
Change-Id: I14666994bc334012a5198a64a4a24cd4962e4bd6
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
|
|
|
|
JIRA: FUEL-205
verify: no-cache
Change-Id: I37746bf3e490f6b707de8fe56af10f1df932908c
Signed-off-by: Michal Skalski <mskalski@mirantis.com>
|
|
|
|
|
|
JIRA: FUEL-205
This change integrates https://github.com/openstack/fuel-plugin-tacker
to the Fuel project.
Change-Id: I51b1bdeeb678e55b85393f6a9f8180fc79322e8f
Signed-off-by: George Paraskevopoulos <geopar@intracom-telecom.com>
|
|
During CI build loops, the workspace is freshly fetched,
so '.docker' make target is built each time.
Reuse existing 'opnfv.org/ubuntu-builder:14.04' tag if present, but
enforce a configurable lifespan for this tag (currently 1 day), so
it does not get stale.
JIRA: FUEL-204
Change-Id: I8d1bd81f3e5e81068e9586c9bea9f43b3327eb38
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
|
|
|
|
When testing with nested virtualzation I attached
the Fuel ISO to VM (/dev/sr0) running deploy.sh
This failed because we expect the ISO to be a file, when
it could be a device, e.g. /dev/sr0. Instead of checking isfile()
check for read access.
This tested with bsdtar. It can extract an ISO from a file or device.
Change-Id: Ic09373cd618b33382303d9134cce4e005040de95
Signed-off-by: Ross Brattain <ross.b.brattain@intel.com>
|
|
pgrep was added to procps-ng, it is more exact
also replace ps -ef in wait_until_fuel_menu_up
removed get_fuel_menu_pid
Change-Id: I57ad9d2fabbfe5f570e0be3bfb40ef842f95902d
Signed-off-by: Ross Brattain <ross.b.brattain@intel.com>
|
|
Change-Id: I706563e6ed1bbfd98adc5a4124ff63789f9e32fa
Signed-off-by: Peter Barabas <peter.barabas@ericsson.com>
|
|
|
|
|