Without the type, the overcloud deployment fails with
Missing parameter type for parameter: CinderNVMeOFTargetProtocol
Change-Id: If9aba2758812d6f8b7370582953653fbfc590ba6
The Cinder NVMeoF deployment YAML has a couple of issues addressed in
this patch:
- Incorrect port is being used. According to the NVM Express specs: "TCP
port 4420 has been assigned for use by NVMe over Fabrics". And that's
the port default in Cinder as well.
- Instead of loading ``nvme-rdma`` we should load ``nvme-fabrics`` so it
will automatically load the right module when connecting (``nvme-tcp``
or ``nvme-rdma``).
- There is no need to load ``nvme`` module, since that's only used for
local nvme volumes, and we care about remote modules.
- Controller nodes also connect to storage (for example for backups), so
we need to load ``nvme-fabrics`` there as well.
- An iptables rule to allow port 4420 is needed, just like the one we
have for iscsi (port 3260).
- Add a new CinderEnableNVMeOFBackend parameter, which triggers
the configuration of the LVM volumes, mirroring the behavior
of CinderEnableIscsiBackend.
Change-Id: I619914a37450dae3dcccbd28e898ca81009bb2bb
This reverts commit 7bba86fc582b6d599117be06c23d0a3d91332cfd.
Reason for revert: Fix regression for virtqemud is started with umask of 0027 which then gets in the way libvirt sets up hugepages. In this particular case, libvirtd/virtqemud creates $hugetlbfs/libvirt/qemu/ (/dev/hugepages/libvirt/qemu/ in this case) which is meant as top level directory, a base where individual, per-guest directories are then created. Now, the top level directory is owned by root:root (so that nobody else can modify its perms) but we also want it to be accessible by other processes (mind you, we don't need to list the directory (o+r), we need to be able to access subdirs (o+x)).
Setting such restrictive umask for a process that is supposed to prepare environment for other users (QEMU runs under qemu:qemu) was a bad choice because of that, hence this revert
Change-Id: I75897fbaf7d9415178bc267c848b83775a384814
Related: rhbz#2166979
Ansible package changed name to ansible-core. Exclude must stay due to
bz https://bugs.launchpad.net/tripleo/+bug/1998501, only update exclude
field to exclude ansible-core instead of ansible.
Related-Bug: #1998501
Change-Id: I2c83d7f494267d4c4bf9e7663a7ece048658a406
Currently rgw_max_attr_size defaults to 256 (a swift default) which
creates problems for openshift on OSP while uploading bigger images.
This patch will update the default value to 1024.
Alternatively, we can enfore a new value to this parameter
using a env file during deployment.
parameters_default:
CephConfigOverrides:
rgw_max_attr_size: <new value>
Change-Id: I9b056e60622a9f2755da6312b89d80577ca596f4
Resolves: rhbz#2167161
The multipathd daemon relies on udev to manage multipath devices,
which in turn requires the daemon have access to /run/udev on the
host. See Ia6a8d27fd2ae6310544bc3767cf7f1fb246939c3 to understand
why we cannot bind all of /run into the container.
Resolves: rhbz#2165494
Change-Id: I46e804676757068e41ebb1e63ba0136d038c0247
This parameter is missing during ovn-controller update and can cause
cut in data plane in certain cases.
We refactor this section of the external_update_tasks to manage all
parameters that have to be set during update for ovn-controller. If
others are needed we just add them to that section.
Change-Id: I30f083a906e457fd5b92d96a6fda8e6451c11fbb
In order to avoid flow update while updating, we need to set
external_ids:ovn-ofctrl-wait-before-clear to some value that will
be used to wait before clearing all flows.
There is code for taking this parameter into account during
deployment, this is the part that takes it into account during update.
Change-Id: I8af36a960156e5203c97293442fbfe1a1ff47dd5
Co-Authored-By: David Hill <davidchill@hotmail.com>
The heat-cfn api uses tcp/13005 for its public endpoint when ssl is
enabled. This corrects the wrong port in iptables rule.
Closes-Bug: #2003929
Change-Id: Ibb8037fad70c628ab26a6cf0dd4401bb6b23cc7c
Make it role-specific for other compute roles, like HCI,
to work with it
Follow-up Ie7ad2d835c1762dc4b9341e305e6a428cb087935
Signed-off-by: Bogdan Dobrelya <bdobreli@redhat.com>
Change-Id: I7d5c01d21af34df5e9fa8842f46aa435975a2ace
When enabling the rsyslog forwarder in tripleo, the amphora logs are not
added to the rsyslog config. Add the 2 files if log offloading is
enabled, and also ensure that the files exist in the directory, rsyslog
skips the settings if they are missing, and they were created only when
a load balancer was spawned by Octavia
wallaby-backport-potential
Change-Id: I04c48371183c84d1c0f4295536c7c3e59483daad
The octavia_rsyslog container fails to start because of a privilige
issue ("could not transfer the specified internal posix capabilities
settings to the kernel, capng_apply=-5"). It appears that the tripleo
rsyslog service is created with "privileged: true", apply the same
parameter to the octavia container.
wallaby-backport-potential
Change-Id: I1e5107c7e2a42e47cb3e14cf6db02482f7b0b331
Closes-Bug: #2003777
qdrouterd for RPC is not a supported backend in RHOSP,
updating sc03 to test rabbitmq which is a supported
backend for RPC communication.
Trying to run with qdrouterd backend in downstream fails
because python-pyngus is not added as dependency downstream[1]
python-pyngus available since RHOSP13 but is not added as dep
for oslo-messaging rpm with reasoning[2].
~~~
ModuleNotFoundError: No module named 'pyngus'
~~~
Sc03 provides desginate testing coverage, This change will
allow us to run standalone sc03 in downstream pipelines keeping
sc03 env files in sync in upstream and downstream.
This patch will be backported to wallaby.
[1] https://github.com/rdo-packages/oslo-messaging-distgit/blob/wallaby-rdo/python-oslo-messaging.spec#L89-L91
[2] https://bugzilla.redhat.com/show_bug.cgi?id=1791364#c13
Change-Id: Ib49223709e20931abcec4d16301db9882c64c06f
"ceph-mon.yaml" should populate collectd hieradata for "ceph_mon"
instead of "ceph_osd".
Add example mons and osds daemons for collectd to monitor in
CI environment files.
Depends-On: https://review.opendev.org/866817
Depends-On: https://review.opendev.org/866688
This change is wallaby-backport-potential.
Resolves: rhbz#2116340
Change-Id: I5d6626774de7aeac9bbc5d70d856eeaba987714b
On EL8 based systems we should use latest container-tools as the
support for running UBI9 is only available there.
Change-Id: I8d777de2778cc595878619552831a1d60086ccef
Ceph plugin in "puppet-collectd" requires Ceph cluster's
fsid[1] to generate the plugin configuration with new socket
paths.
This change comes after the socket path of services were
changed from "/var/run/ceph/" to "/var/run/ceph/<fsid>/".
This change is wallaby-backport-potential.
[1] https://github.com/voxpupuli/puppet-collectd/pull/1007
Signed-off-by: Yadnesh Kulkarni <ykulkarn@redhat.com>
Change-Id: Ic67f68ec51e176040ef982a3721cd7afce68877e
