| Andreas Pflug 2005-12-29, 7:24 am |
| Andreas Pflug wrote:
> Is the algorithm to conduct a failover considered stable over versions
> (subject to backward compatible improvements), or do we need a specific
> slonik suitable for each cluster version?
>
>
> For the next pgAdmin release, I'd like to have slony failover support
> included. Failover is performed in a situation where probably bad things
> already happened, and bad things seed hectic actions that may lead to
> even more bad things. To support admins in keeping cool, I'd like to
> provide them a tool that checks the most critical issues.
>
> I'm planning two tools: an upfront health checker, and the failover tool
> itself.
>
> The health checker will check if I have all information, i.e. connect
> strings for the admin's workstation at hand, in case I'll need them.
> Could be quite a nightmare if you accidentially mix up your test server
> with a live replication node...
>
> The second tool will do essentially what slonik failover does. Most
> other slonik functions are reimplemented in pgAdmin itself, and for
> failover I'm wondering if that's viable too or not. Thus the question
> above.
>
> I'd prefer to implement the failover algorithm in pgAdmin itself, but if
> necessary I'd assemble parameters for slonik and call it.
Hm, I never received an answer on this. Could one of the slony core
developers give me some advise?
Regards,
Andreas
|