[Linux-HA] Feedback: Master/Slave RA for Postgres / Slony Cluster?

[Carlos Alarcon]

Searching the LINUX-HA list archives I have found this thread.
I was actually looking for an implementation of postgresql ocf RA which
supports slony-I master/slave behaviour.

My current needs are only for two nodes ( simple master/slave schema) with
failover support.

Did you finally implement any of the discussed functionality?

thanks



On Nov 7, 2007, at 10:19 AM, Dominik Klein wrote:

>> Feel free to add it as an enhancement request in bugzilla - i think  
>> it makes a certain amount of sense (in general, not just in your  
>> case) to implement
>
> Ok.
>
>>> Read my first post again.
>> I read it a few times but found myself getting lost - maybe i was  
>> just having a bad day :-)
>
> No problem. I was pretty sure this had to be read several times, but  
> I could not make it any clearer. I tried though .. :)
>
>>> 1. monitor on A fails -> OCF_NOT_RUNNING
>>> Now, instead of "demote A, promote B":
>>> 2. Stop/Start the resource on A
>> One can certainly have demote->stop->start->promote (all on A)
>> Does that help?  The demote shouldn't be a problem for you because  
>> its a no-op.
>
> Would help.
> Demote on the not running resource would succeed, stop too. Start  
> would propably succeed and promote would result in "master already;  
> return SUCCESS"
>
> Is this a configuration thing or would that have to be implemented  
> in Linux-HA?

Configuration.
By the sounds of it, you just need failure stickiness to be zero for  
that resource.

>
>
>> The notify data will also tell you that you used to be a master and  
>> that no-one else is one at the moment... in theory that should be  
>> enough info to allow you to do something clever right?
>
> I never dealt with the notify stuff yet. I will have a look at it.

http://linux-ha.org/v2/Concepts/MultiState#head-6c46b6ac3329cf29a3afb8eb7d54
9007b0d52311

>
>
>>> Iirc "start" includes a monitor action (or "probe" called  
>>> sometimes in this case).
>> Not really... we probe only when we dont know the current state of  
>> a resource on a node.
>> So generally this only happens at startup or when the admin uses  
>> crm_resource -C
>
> Ok. That's why i wrote "iirc" which could easily be false :)
>
> Regards
> Dominik
> _______________________________________________
> Linux-HA mailing list
> Linux-HA at lists.linux-ha.org
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems