Age | Commit message (Collapse) | Author | Files | Lines |
|
* align dpdk packages list
* add missing libvirtd group to vcp host nodes
Change-Id: Ideab2b16ca0561035b225e58bf3d0c5653d303bb
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
* up glusterfs.client state before actual volume usage
* handle keystone server state
* specify suitable nova packages list for uca repo
(consoleproxy vs novncproxy)
* upgrade vcp nodes to get proper cryptography
library for keystone
* align service names for libvirt & glusterfs
Change-Id: Iaeb7d147e6d407bbeaec2d40fd81037c939c3fe0
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
Currently, Xenial repos provide MaaS 2.2.x, while the PPA bumped it
to 2.3.x. Since we switched to 2.3, we observed a rare wrongful state
transition from 'Deploying' back to 'Ready'.
Drop the PPA, falling back to 2.2 from mainline distro repos.
JIRA: FUEL-312
Change-Id: I3daa118059f37cbeca076da685661c28f3a28a97
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
To align with new cluster naming convention, rename 'openstack'
state file to 'openstack_noha'.
While at it, factor out baremetal setup from 'virtual_control_plane'
into a new state that will be reused in upcoming scenarios, remove
useless sync_all (automatically done after node reboot).
FUEL-310
Change-Id: I6d7e5db8f09305f2fd8eeca0199a2e85b08d2202
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Change-Id: I879f565fe17539a54770fa22a6088c6493524e0c
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
Although we add default routes via public network and disable
DHCP client from setting new routes, until we reboot the prx* nodes
we still have the stale route originally set at initial boot.
Change-Id: Ib8e5fb67c7da00684e0ac21984fc4661d3820d83
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Change-Id: I1df0228cb44bf9122aaf93dd25fc16a0d26a5240
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
|
|
Occasionally, cinderng.volume_type_present errors with:
ClientException: Service Unavailable (HTTP 503)
Instead of retrying the whole state file, use `wait_for` macro
to retry only this high state up to 5 times.
Change-Id: Ib9ef017aca737e53c853007c13107d56d856c016
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Upstream fixed the salt-formula-horizon in commit 95387ec, by
defining 8078 (and only that) port in Apache's ports.conf.
This fixes the port 80 overlap, so running the `apache` high state
after the `horizon` high state not only is unnecessary now, but
also would lead to new breakage, since `apache` state would
overwrite the ports.conf (removing 8078 and adding 80), i.e.
creating a new port conflict and breaking Horizon port completely.
This reverts commit eb4645206d6d74992fca3b8726ee2eebca97205f.
Conflicts:
mcp/config/states/openstack_ha
mcp/reclass/classes/cluster/baremetal-mcp-ocata-common/openstack_proxy.yml
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
Change-Id: Iea8f0bd90ee8d12f399aad16247dda274d6a907a
|
|
Employ UCA repo on computes nodes for nosdn-nofeature-ha scenario as
well to prevent a regression (creation of ports failed for 1+n instances)
of neutron ovs agent from mcp/openstack repos.
Change-Id: Ie65ae122096c0d3a93c09d46191787a934bd7d4f
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
|
|
In order to complete broken database sync run
nova state on controller one more time.
Change-Id: I761f26667ebb531b848a62e096f3d79f588d9f24
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
NOTE: In order to undefine VCP VMs with NVRAM (e.g. AArch64 VMs
using AAVMF), an additional parameter should be passed to libvirt
by Salt virt core module (equivalent to `virsh undefine --nvram`).
While at it, pass CI_DEBUG, ERASE_ENV enviroment variables to
state execution, and stop force-applying patches.
Also refactor the rsync between foundation node and Salt master,
so the whole git repo is copied as </root/opnfv>, and <root/fuel>
becomes a link to it; useful for Armband, where 'fuel' is a git
submodule. Fix .git paths after rsync, so git submodules work as
expected in cfg01 repos.
JIRA: FUEL-307
Change-Id: Ic62f03e786581c019168c50ccc50107238021d7f
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
- prx: add route for public traffic to public interface
- prx: add route towards salt master through maas
- remove dashboard class from proxy node (already implements horizon)
- remove dashboard (and benchmark) class definitions (no longer used)
- (temporary) backport Pharos change for adapter template
JIRA: FUEL-305
Change-Id: Ia14a18ac0123c1134d8d99dc43da9a1f770001d0
Signed-off-by: Guillermo Herrero <guillermo.herrero@enea.com>
|
|
|
|
|
|
JIRA: FUEL-306
Change-Id: I648545890c1f7dc59176beac1a0593aed54079cb
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
Signed-off-by: Delia Popescu <delia.popescu@enea.com>
|
|
Generate documentation automatically using `reclass-doc`.
nginx is already configured to serve said documentation on
proxy's public VIP on port 8090.
Change-Id: If2aef646a0ec44d5cc7e9d425e565e5c0aa581b3
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
JIRA: FUEL-301
Change-Id: Id6b2b423b8045c581fa5c02133cf91702d9915c9
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Apache module will take care of ports.conf file to prevent
bind socket conflict between apache & nginx services.
Change-Id: Ia76ec356002e1db0dabd20d8f355a1b16fc07b30
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
Change-Id: Ic81507f3f7b3fec593b507e0c534434e8489b01b
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
JIRA: FUEL-301
Change-Id: I9de98fb961fd1d480b45a774de61ad6a93e9addf
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Change-Id: Icc30d27951abb1e231c9269c6293782a39e08fb6
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
|
|
Since VMs are not affected by this limitation, only apply the fixup
to baremetal nodes.
JIRA: FUEL-299
Change-Id: Ib94c481627538d900295df03b8c8fdc7b61cd718
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Change-Id: Ia896c3f9fcd96dd498eef6d1f83d46e29df0cd6b
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
Apply aodh state in consecutive order to avoid
a race condition with database synchronization.
Change-Id: I4684fbeaaba2c9780084e0a64fe6453bccfb67e0
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
* neutron on computes (dpdk case: void state)
* mongodb server (incomplete initialization)
Change-Id: I3dd3266b5c2d1b155981f725e15742cd38ed899d
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
Change-Id: I95c284cbf374194694360bffbeaf6770db6111bf
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Change-Id: I7a21c30d49aecca948f45535fec164c2f643450e
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
The recent addition of `linux.system`, combined with `system.reboot`
for the baremetal compute nodes leaves compute nodes unconfigured
after reboot.
Run `system.network` too, but expect a failure (only for DPDK, which
requires hugepages to be already active, hence a prior reboot).
Fixes: 64920b8
Change-Id: I8c73b24ae15e1f87dee64ae2aba7af86db1e942f
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
After MaaS reports baremetal provisioning finished successfully,
check that all nodes are online before attempting a `sync_all`.
Change-Id: I6ba4b3e4ba5b5258ace4da8c39e0fc77354885e3
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Change-Id: Ib4aa3f2cb4fc7129d502b4332cd7fedd83a0e1fe
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
DPDK scenario requires hugepages to be set up and enabled before
configuring OVS ports, so apply `linux.system` on compute nodes
and reboot them before attempting to run `linux.network` state.
Change-Id: I2b685361b07727a4740a3676c5f219a6443d1cf2
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
In order to set properly keystone fernet keys, apply glusterfs client
state before second keystone server state. Also leave out user/group
settings for glusterfs volume of nova instances as it will be set later
by nova compute packages themselves.
Change-Id: I069e37c67f08c51ed29f45cf6f92d4a00a1ac97b
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
While applying scenario states, break on error, and retry failed
state up to 5 times. Apply the same behavior for `salt.sh`.
Add new deploy parameter, '-D', backed up by 'CI_DEBUG' env var,
which gates deploy sh scripts logging (set -x).
Also extend '-f' deploy parameter, allowing it to be specified
more than once; the first occurence will skip infra VM creation,
but still sync reclass & other config from local repo, while a
second occurence will also disable config sync.
To prevent glusterfs client state from failing due to non-existent
nova user/group, move it after nova:compute's nova state is applied.
Change-Id: I234e126e16be0e133d878957bd88fed946955de8
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
While at it, compact 'set' into bash shebang where possible and
add `make patches-copyright` target to simplify adding patch
license headers.
Change-Id: I0c841de72e5709e5eef915a52c5ec4a7fc0f7c37
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
While at it, fix some shellcheck warnings, and s/fgrep/grep -F/g.
Change-Id: I093b7b4c196731b1ecc0c27a4111955b2e412762
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
We should eventually also support baremetal deploys without a
virtualized control plane (VCP), so decouple MaaS provisioning
from VCP provisioning.
While at it, move "wait_for" bash function from maas state to
common library file, lib.sh.
Change-Id: I32c33135655cb6aceae901a5f92b51265a8c84b4
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
* use pseudo agentdb port binding controller instead of
the deprecated network topology one
* disable superfluous l2population mechanism driver
* tidy up the duplicated haproxy neutron listen opts
* straighten karaf features list
* update jetty config
Change-Id: Ifacf8de11eb56ab72df13a312151a510b280dea2
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
Change-Id: If7cb8473f5c290d1d5f22fce5567f7b8da24fd9f
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
Test of snaps_health_check gets console output of instance, which is
empty (https://bugs.launchpad.net/ubuntu/+source/libvirt/+bug/1667033),
therefore take affected packages of computes from UCA repo.
Change-Id: I4e13a40dd47caf305efd9f393b3c3cf5a17ef312
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
|
|
Change-Id: I53ac0be519df1bb39a6a56e236285fce95228bd4
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
Change-Id: I7fe8d0c77a1d62e2214fb1089651a639303dd20e
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
pkg.upgrade was enabled for all salt minions, including VCP VMs,
which take longer to perform the operation, probably due to an
older set of packages in the Ubuntu disk image we use.
One way to work around this is to switch to UCA Xenial image, and
let Salt pre-provising salt minion on it, but that adds deploy
time delay and has caused issues in the past (should be ok now).
Alternatively, we can retry the pkg.upgrade until all minions
respond, before moving on with the state execution. This prevents
silently skipping the next salt calls (e.g. installing keepalived).
Note that the issue did not manifest for OVD-DPDK, where after
pkg.upgrade, DPDK is installed, giving VCP VMs enough time to return.
While at it, retry 'salt.control' state apply too (non-critical,
but it fails every once in a while).
Fixes: 87310fb
Change-Id: I97acc2b23206a55d72f7e6583ca42127fdbacc16
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
Occasionaly, MaaS fails to provision/deploy some nodes, in which
case we try marking them as broken, then fixed (to put them again
in 'ready' state); before re-attempting the MaaS deploy operation.
However, this leads to 'Error: Internal server error' when deploy
function is called right after transitioning the node to 'ready'
state.
Add a delay of 30 seconds before re-attempting the failed operation.
Change-Id: Ia9ecec67639387e4a29feab3114e1741c554a2cb
Signed-off-by: Alexandru Avadanii <Alexandru.Avadanii@enea.com>
|
|
* [baremetal] add memory to contollers & salt master
* tune up sysctl vm.dirty* for compute nodes
* upgrade packages to get the latest versions
(https://bugs.launchpad.net/cinder/+bug/1641312)
Change-Id: I9ad22206f2f3f11e1da3f93c7a0931c592adf1cf
Signed-off-by: Michael Polenchuk <mpolenchuk@mirantis.com>
|
|
|