On the topic of contribution for masakari, etc.    I have made a decent amount of changes to my masakari platform in terms of how pacemaker reports in, notably I run pacemaker on each node and not the pacemaker-remote.    Haven’t figured out a better approach yet.   I have also rejuvenated the watcher ansible project in my labs    Would love to contribute in some ways.   

Get Outlook for iOS
From: Dmitriy Rabotyagov <noonedeadpunk@gmail.com>
Sent: Wednesday, June 19, 2024 11:30:32 AM
To: laszlo.budai@gmail.com <laszlo.budai@gmail.com>
Cc: dwest576@me.com <dwest576@me.com>; openstack-discuss@lists.openstack.org <openstack-discuss@lists.openstack.org>
Subject: Re: [openstack-ansible] [masakari]
 

Um, I'm not sure if VM context is applicable here at all.

So pacemaker/corosync cluster is used by masakari-hostmonitor to independently identify that some host is down.
It achieves that by listening for corosync broadcast messages on the interface. Once the masakari monitor detects that the host is down (by seeing the traffic saying that), it creates an evacuate event in masakari API, and then masakari engine issues a host evacuate request towards nova API.

Regarding stonith, I know ppl made it, but I personally never played with it yet, so can't help much now.
But all contributions to roles functionality are warmly welcome, so if you spot place for improvement or missing functionality - don't hesitate to bring that up:)


On Wed, Jun 19, 2024, 20:21 Budai Laszlo <laszlo.budai@gmail.com> wrote:
Do you have any STONITH agent configured? Or are you using fabric fencing?
The doc mention that fencing should be configured to make sure that an unreachable but possibly still running instance should not be able to access any of the data of the VM, so the replacement VM can safely access the data.

Thank you.


On 19.06.2024 19:30, dwest576@me.com wrote:

My crm_mon also shows no resources, I think it is normal.

 

Fencing is done through creating segments, then adding hosts to said segments.   My fencing mimics my availability zones.

 

From: Budai Laszlo <laszlo.budai@gmail.com>
Sent: Wednesday, June 19, 2024 9:09 AM
To: OpenStack Discuss <openstack-discuss@lists.openstack.org>
Subject: [openstack-ansible] [masakari]

 

Dear All,

I'm evaluating Masakari for instance HA and I need to install it using openstack-ansible. I'm doing it initially in a small test cluster (3 controller, and 3 compute).

The install went trough and now when I'm checking the status of the corosync cluster I see the following:

root@worker1:~# crm_mon -1
Cluster Summary:
  * Stack: corosync
  * Current DC: worker2 (version 2.1.2-ada5c3b36e2) - partition with quorum
  * Last updated: Wed Jun 19 16:02:01 2024
  * Last change:  Wed Jun 19 15:29:37 2024 by hacluster via crmd on worker2
  * 3 nodes configured
  * 0 resource instances configured

Node List:
  * Online: [ worker1 worker2 worker3 ]

Active Resources:
  * No active resources
root@worker1:~#

is it normal that there are no resources configured? What about the fencing, how should I configure it? Are you aware of any documentation where I could find mori info about configuring fencing (especially in the context of openstack-ansible)?

Thank you for any advice.

 

Kind regards,
Laszlo