[Pike] when Gnocchi used, crm complains that haproxy isn't installed

[Xav Paice]

Fresh cluster with 18.02 charms, Juju 2.3.8, and Xenial/Pike.

On a ceilometer unit:

crm status
Last updated: Tue May 29 04:32:44 2018 Last change: Tue May 29 04:32:41 2018 by hacluster via crmd on juju-2fe2c0-8-lxd-1
Stack: corosync
Current DC: juju-2fe2c0-8-lxd-1 (version 1.1.14-70404b0) - partition with quorum
3 nodes and 6 resources configured

Online: [ juju-2fe2c0-28-lxd-3 juju-2fe2c0-6-lxd-1 juju-2fe2c0-8-lxd-1 ]

Full list of resources:

 res_ceilometer_agent_central (lsb:ceilometer-agent-central): Started juju-2fe2c0-8-lxd-1
 Resource Group: grp_ceilometer_vips
     res_ceilometer_eth0_vip (ocf::heartbeat:IPaddr2): Started juju-2fe2c0-6-lxd-1
     res_ceilometer_eth1_vip (ocf::heartbeat:IPaddr2): Started juju-2fe2c0-6-lxd-1
 Clone Set: cl_ceilometer_haproxy [res_ceilometer_haproxy]
     Stopped: [ juju-2fe2c0-28-lxd-3 juju-2fe2c0-6-lxd-1 juju-2fe2c0-8-lxd-1 ]

Failed Actions:
* res_ceilometer_haproxy_monitor_0 on juju-2fe2c0-8-lxd-1 'not installed' (5): call=83, status=Not installed, exitreason='none',
    last-rc-change='Mon May 28 22:04:19 2018', queued=0ms, exec=0ms
* res_ceilometer_haproxy_monitor_0 on juju-2fe2c0-6-lxd-1 'not installed' (5): call=38, status=Not installed, exitreason='none',
    last-rc-change='Mon May 28 21:57:48 2018', queued=0ms, exec=0ms
* res_ceilometer_haproxy_monitor_0 on juju-2fe2c0-28-lxd-3 'not installed' (5): call=57, status=Not installed, exitreason='none',
    last-rc-change='Mon May 28 22:04:19 2018', queued=0ms, exec=0ms

It's totally right, it's not installed, and isn't needed.

[James Page (james-page)]

This is related to the removal of the ceilometer-api at Pike/Queens.

[James Page (james-page)]

infact there is not much point in having hacluster deployed at-all in later releases - ceilometer is just middleware that sends and receives messages from RMQ - i.e. no VIP is needed, nothing to access.

[Xav Paice (xavpaice)]

Given that we need the ceilometer_agent_central to be managed by pacemaker, can we tweak the charm to just not create the VIP and haproxy resources?

[Drew Freiberger (afreiberger)]

Is this being caused because we still set a vip for an api that no longer exists?

[Xav Paice (xavpaice)]

The haproxy resource is still being created, but haproxy isn't even installed (19.04, Bionic). The only way to get clean monitoring for this is to manually delete the resource from Pacemaker, by removing the lines:

primitive res_ceilometer_haproxy lsb:haproxy \
        op monitor interval=5s

[Chris MacNaughton (chris.macnaughton)]

This, unfortunately, has nothing to do with the VIp and everything to do with an assumption that was made in ancient history that every charm using hacluster had haproxy running. That needs a change in charm-helpers and to be synced into ceilometer.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.opendev.org/671523

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.opendev.org/671523
Committed: https://git.openstack.org/cgit/openstack/charm-ceilometer/commit/?id=7b523f82cd139c83cf5a39335a112403d8c3cbfd
Submitter: Zuul
Branch: master

commit 7b523f82cd139c83cf5a39335a112403d8c3cbfd
Author: Dan Ackerson <email address hidden>
Date: Thu Jul 18 18:30:36 2019 +0200

    Disable haproxy in ocata and beyond

    Also take care to remove haproxy in case of upgrades.

    Change-Id: I9dfdfc366fb64e50f0209e85072f66998a69f0ac
    Closes-Bug: #1773894

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/19.04
Review: https://review.opendev.org/674676

[OpenStack Infra (hudson-openstack)]

Change abandoned by Paul Goins (<email address hidden>) on branch: stable/19.04
Review: https://review.opendev.org/674676

[Drew Freiberger (afreiberger)]

This was fix-committed against stable/19.07 at commit:
https://github.com/openstack/charm-ceilometer/commit/7b523f82cd139c83cf5a39335a112403d8c3cbfd