Hi guys,
I`m in the process of implementation of Ansible playbooks for Openstack/Ceph rollout. As part of our requirements, we have HA to pretty much anything.
I believe, that HA is not applicable for Ansible, since it`s just opens an SSH connection, and does commands on remote site.
Have anyone thought about this? Maybe some approaches or best practices - I sniffed around, but nothin so far.
Cheers,
NM
Hi guys,
I`m in the process of implementation of Ansible playbooks for Openstack/Ceph rollout. As part of our requirements, we have HA to pretty much anything.
I believe, that HA is not applicable for Ansible, since it`s just opens an SSH connection, and does commands on remote site.
Have anyone thought about this? Maybe some approaches or best practices - I sniffed around, but nothin so far.
Cheers,
NM--
You received this message because you are subscribed to the Google Groups "Ansible Project" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ansible-project+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
You could consider running ansible in pull mode, on each node, and retrieving updated playbooks etc via `git fetch --all` with 2 remotes (eg github and bitbucket, or your own gitolite servers).
Not sure how "HA" that is but it works fine for me.
MfG/Cheers
Dave Cottlehuber
Nick,
Nick Maslov schrieb (07.10.2013 14:38 Uhr):
I`m in the process of implementation of Ansible playbooks for Openstack/Ceph rollout. As part of our requirements, we have HA to pretty much anything.
I believe, that HA is not applicable for Ansible, since it`s just opens an SSH connection, and does commands on remote site.
Have anyone thought about this? Maybe some approaches or best practices - I sniffed around, but nothin so far.
I'm not sure what exactly your problem is.
Is it that you do not want to interact with (files/configuration of) a service which is controlled by a Cluster Resource Manager (CRM) and should not be stopped/restarted by ansible bypassing the CRM?
Marc
Nick,
Nick Maslov schrieb (07.10.2013 14:38 Uhr):
>I`m in the process of implementation of Ansible playbooks for Openstack/Ceph rollout. As part of our requirements, we have HA to pretty much anything.
>
>I believe, that HA is not applicable for Ansible, since it`s just opens an SSH connection, and does commands on remote site.
>
>Have anyone thought about this? Maybe some approaches or best practices - I sniffed around, but nothin so far.
I'm not sure what exactly your problem is.
Is it that you do not want to interact with (files/configuration of)
a service which is controlled by a Cluster Resource Manager (CRM)
and should not be stopped/restarted by ansible bypassing the CRM?
And something else to think about is:
To not bring down all systems in a HA-cluster at the same time (rolling upgrades).
git and multiple checkouts is how i do HA of ansible itself
Hi Brian,
Yes, this is something closer to what I`m thinking about. Having ansible in crontabs on local servers is also an approach.
Have you considered using HAProxy in front of few Ansible servers, that are feeding from git/svn with fresh playbooks/inventory files?
That will require you to force outbound connections to use HAProxy VIP to minimize amount of SSH keys pushed to managed servers (one instead of equal to the number of servers). Any idea re this approach?
Running ansible playbooks via Python API or direct ssh command on HA VIP is also something I`m trying to achieve.
Cheers,
NM
Generally speaking, ansible is naturally decentralized and serverless, so it would just be a matter of having multiple copies of the control machine.
If you’re looking for making periodic remediation HA, you could use AWX callbacks so the systems phone home to a particular address and have a replica available.