After libvirt pod restarts, the virtual machines created before
turn transient ,then opetrate these vms ,nova-compute will throw
exception. This is because that the directory /etc/libivrt/qemu
in pod contains vitual machines xml files and it is temporary,
the xmls files disappear after the pod restarts, so we mount it
to hostpath /etc/libvirt/qemu.
Change-Id: I48fd712c2b0565cb2cfe850482e8501f4e5022a4
Closed-bug: 1760003
This PS updates the heat templates, reducing the size of the
launched vm. In addition cinder is enabled in the ceph dev
pipeline, this is possible due to the resources no longer
consumed by the test vm.
Change-Id: I9efe6fe643c636b660dd54b60dfe7c8785d7fec2
This PS disables the multinode ubuntu gate from voting, which has
been failing due to -infra issues - and severly hampering development
work as a result.
Change-Id: I411ebe20ba19c52475b559952712faf742343673
This PS moves all the config files to be directly values driven,
both simplifying over-ride and allowing configs to be targeted
to pods in future work.
Change-Id: I7e16585c9ef49275327d19a48f00bad192dc4923
This PS gives ingress rules attached to the cluster wide ingress
controller the suffix `-fqdn` to allow them to be used.
Change-Id: I7de85e349fb609b8380070030579b9b4767e72d1
This PS moves all the config files to be directly values driven,
both simplifying over-ride and allowing configs to be targeted
to pods in future work.
Change-Id: I286af7434aab6de941f9700a7fbf70c6dd0ee4cb
This PS moves all the config files to be directly values driven,
both simplifying over-ride and allowing configs to be targeted
to pods in future work.
Change-Id: Ifcbc19b17aa1d145f12ed1aed8b15a69ca045bb7
This PS udpates the frequency and initial delay on the mons
livenessProbe, to allow time for the cluster to restart if mons
get into a crashloop backoff following power outage.
Change-Id: Iea74c4d52882a157a84f4f12bc411f2014869f99
Recent versions of MariaDB (10.1.31, 10.2.13) have a regression that
breaks clustering. See https://github.com/MariaDB/server/pull/457 and
4e6dab94d0
for an in depth explanation.
We need 10.2.13+ for Barbican to function correctly (see bug #1734329)
but we also need the fix above to support MariaDB clustering.
This work-around can be removed later on when MariaDB 10.2.x releases
contain the needed script fix.
Thanks to Sam Yample <sam@yaple.net> for helping track this down.
Change-Id: Ifd09d7effe7d382074ca9e6678df36bdd4bce0af
This PS adds the ablity to target different configs to different
hosts for the nutron agents, in the same manner as nova-compute.
Change-Id: Iebd87e30014d6cac2127c7e1a14259b10d74fbf8
This reverts commit 81bf5f3656 back to
MariaDB 10.1.23 which we know works with clustering enabled
(pod.replicas.server > 1).
Change-Id: Ibf70dbab78f03d32e1ec96e99ac8db59d23cb96e
This PS allows arbitary hostnames to be used for public endpoints,
provided the resolve externally to the ingress controllers.
Change-Id: I44411687f756968d00178d487af66c2393e6bde0
This version is already being used by some charts, so this brings the
rest of the charts in line and allows them to use a new feature,
pod dependencies, that this version provides.
Change-Id: Ie8289eb09b31cd8f98c2c5b4dd5bbe469078e6d8
