In other words when using --ssl, if any service is explicitly
requested to be HA don't override it with the single unit
HA hack that we use to get predictable SANs for ssl certs.
If we are able to glean a dns server from the local
host, don't both asking for one as input since this
happens for almost every openstack deployment and is
quite tedious and unnecessary.
Reduce default log level for Ceph template and
related overlays.
Log level 20 is highly verbose and fills up the log files
pretty quickly. With loglvel=20, osd log files
(/var/log/ceph/ceph-osd.*.log) have reached ~20G in ~8
hours.
Reduce loglevels down to 1 based on Dan's(hillpd) comment:
"Ceph supports run-time logging level changes. It is easy
enough to use `ceph tell` or the daemon socket to
raise/lower log levels across a reproducer.
I'd recommend dropping the level all the way down to 1 by
default.
This will also reduce overall stress on stsstack -- ceph
can be very verbose at raised debug levels."
Don't exclude namespaces charms from revision-info
Most charms release to cs:<charm> as their stable whereas
others use cs:~<team>/<charm> so we need to allow that
when using --release-info.
One remaining caveat here is that if the revision-info uses
a namespaced charm but the template does not then it will
not be matched. That will either require a manual edit to
fix (--replay) or we need to update the template in question
to use the namespaced charm.
We currently expect charm version info given to
--revision-info to be for stable charms and as
a result we silently ignore any charm that is
not cs:<charm>. This patch displays a message
showing any charms whose version is ingored
and if no versions are found, does not switch
to --use-stable-charms.