On 10/06/2014 12:13 PM, Glyn Astill wrote:
>
>>
>
> Thanks Steve.
>
> Yeah this same scenario in 2.2 works fine, but 2.1 is looking pretty broken to me in terms of failing over at all if another node is down in addition to the origin.  Reordering the resubscribe of node 4 after the failover doesn't work, but also just attempting to failover fails when it can't contact the non-origin failed node; it appears slonik want's to connect to the other downed node regardless complaining "ERROR: no admin conninfo for node 3" if I remove the conninfo or failing when it can't connect when I leave it in.
>
> Was this something that was always impossible, or was it canges in 2.0/2.1 that introduced the issues?  I don't recall having the same issue in 1.2 - but it's a while ago now since I checked.
>

You could try disabling the automatic wait for stuff with the -w option 
to slonik to see if it makes a difference.