[ClusterLabs] Pacemaker/Corosync on FreeBSD

[Alberto Mijares]

On Mon, Dec 4, 2017 at 6:53 PM, Jan Pokorný <jpokorny at redhat.com> wrote:
> Hello Alberto,
>

Hello Jan,

Thanks for your feedback.


> On 04/12/17 16:12 -0400, Alberto Mijares wrote:
>> At this point, I need to know if someone is using pacemaker/corosync
>> on FreeBSD. Is it a problem with crmsh only?
>
> well, it's enough to have a look at which people develop these high
> level tooling (crm, pcs) and you'll figure out that these are deeply
> entrenched in linux world, and that's not by accident.


To be honest, I don't believe very much in portability. It should be
just like this.


> And by
> definition of being sort of an assistants to guide the cluster
> configuration, it's natural they try to prevent whatever
> misconfigurations, even beyond pure cluster domain, they deem
> important.  I think tailoring for whatever other platform is still
> possible, but someone has to actually come up with patches and, even
> better, offer an active maintenance of those bits going forward, as
> they are not going to get proper attention otherwise, and unmaintained
> chunks of conditionalized code are usually worse than none.
>

I'm cloning the repo and giving it a try. If someone can point me to
the files/dirs where I should put my effort, I'd appreciate it.

> On the other hand, the stack's core should should play fairly well
> with POSIX-leaning systems, so please, do not feel discouraged, you
> can still setup corosync.conf by hand, use low level tools of
> pacemaker like cibadmin (+ manual edit of the CIB configuration),
> crm_resource and crm_mon, go this lesser convenient way (subject
> of individual evaluation), and still be happy with that.

I can do that. I'll try to make it fun.

>
>> If I configure everything by hand (no crmsh nor pcsd) should it
>> work?
>
> Definitely (and if not, we want to know).

I'll let you know, for sure.

>
> Note that there are some apparent limitation outside of linux,
> like no support for systemd-based services natively managed with
> pacemaker, but that's to be expected, right?

Not only expected but desired (we don't want to do anything with systemd ;-)


Bests...


Alberto Mijares