machine#0 jujud using ~100% cpu, slow to update units state

[JuanJo Ciarlante]

FYI this is an openstack nonHA deployment, single juju state server (ie no ensure-availability)
juju-agent 1.23.3.1

* at machine#0, excessive CPU and RSS usage by juju (and mongod):

ubuntu@node0:~$ ps -eopid,pcpu,rss,vsz,comm,start_time --sort=-rss|head -4
  PID %CPU RSS VSZ COMMAND START
13113 94.6 3174452 8474916 jujud Jul13
 4757 58.5 705932 8850824 mongod Jun04
13690 0.0 369028 1349976 rsyslogd Jul13

* trying a dummy deplloyment (cs:ubuntu) smooshed against an existing
unit takes long time (~8mins) to get updated state to be shown by juju status,
see http://paste.ubuntu.com/11922260/.

Note this is a production environment, if you need more logs feel free
to PM me directly.

[Alexis Bruemmer (alexis-bruemmer)]

Can you please repo this issue on 1.24.3.

[JuanJo Ciarlante (jjo)]

Unfortunately this is a production critical environment, which
I find hard to upgrade unless we have:

1) a stable 1.24.x production release (we've reported several
bugs while deploying our stacks with 1.24.x, iirc related to
leader election fixes)

2) a devs' backed / CI'd 1.23.3.1 => 1.24.x upgrade path.