Question Failover Clusters

Rob Fitzpatrick

Rob Fitzpatrick

ProgressTalk.com Sponsor
I've had an inquiry from a client about Failover Clusters. Looking for input from anyone who has hands-on experience with it.

The client is on AIX 7.1 and would likely be deploying with a mid-11.x release. They use TDE today and would also likely implement OE Repl Plus at the same time.
 
HACMP works very well but the support that OE has for it is marginal at best.. A cluster solution is not really a cluster solution without multiple mirrored databases live at the same time.

OE Repl plus isn't quite ready for prime time imo... there are a number of issues with quick failovers and the ability to keep the source database up when the target goes down as well as the need to bounce the target database to allow users write access (and to restart it as a source database).

Some of the issues are resolved in 11.6... but most of the issues haven't been addressed. Hopefully in future releases.

It is however better than nothing and marginally better than roll your own AI replication because of the streaming nature of replication. Mostly because you are able to keep your warm database much closer to real time.

Async is the only real option because synchronous replication is just way too slow for any realistic data volumes. Testing simple transactions in synchronous mode added almost a second to a simple create of a basically empty record.

TDE - will add basically no extra steps or overhead to the process other than making sure the dbname.ks files are populated where they need to be.
 
Worth getting in touch with Paul Koufalis and Gilles Querret. They did a workshop at the EMEA PUG in 2014 about high availability and Gilles talked a lot about clusters etc. As it wasn't relevant to me I didn't make too many mental notes.
 
I attended that workshop at PUG Challenge Americas 2014, though I don't remember discussion of Clusters. I'll look through my notes; thanks.
 
Rob Fitzpatrick said:
I attended that workshop at PUG Challenge Americas 2014, though I don't remember discussion of Clusters. I'll look through my notes; thanks.
Click to expand...
Just been thinking, and maybe it wasn't Gilles who was talking about that. It could have been Gareth Payne at Progress UK.
 
After talking to someone about this, yes it was Gareth Payne. And he has subsequently tried to recreate that and failed after spending a long time at the job.
 
You must log in or register to reply here.
Back
Top