Active/Active Cluster Use for Failover

[Tim Malaney]

We are currently running Cloverleaf 6.0.2 in a HA clustered Linux environment. The cluster and both servers are located in the same physical site. For disaster recovery, lose of the current site, we are looking at other means of failing over Cloverleaf to a different physical location.  I have seen the term Active/Active Cluster thrown around as another means of performing a failover. To be honest, I don’t understand the concept. Sounds like all threads are using the Local Binding Address option but don’t understand how that works. Is anyone utilizing this in their failover configuration and if so, can point me to some detail on it’s setup and use?

Thanks

Topic

[Tim Pancost]

I believe Active-Active is just a type of HA environment where both machines are “active”, in that they are in use(either as production or non-production).  Either machine can failover to the other, with each machine being capable of running the workload of both, if temporarily.  In an Active-Passive model, the “passive” machine is really just a shell, and isn’t actively running anything on it.  It’s just waiting for something to happen to the primary machine, and it will take over the workload.  This is the model we have here for HA.

As for DR, where you want to run your engine in a geographically distant location in addition to having a local capability for HA, I’m sure there are a plethora of technical options that would be specific to your environment/needs.  For us, we have separate DR machines in another data center several states away.  We installed Cloverleaf on them.  We then developed custom shell scripts to tar up the sites on our productions servers and SCP them down to the DR boxes, via crontab.  There is then another cronjob that runs on the DR boxes to expand the tar files out.  This happens once a day, in the early morning hours.  So, we end up with a copy of our production Cloverleaf sites in a different data center that are, at most, one day “out of sync”.  That is, any changes that would’ve been made that day would be lost in the event of a disaster.  To us, a small price to pay given that such and event would probably mean the destruction/disablement of the primary datacenter, which is a far bigger deal than losing a table entry or Xlate change, or even a new interface.

If such an event were to occur, we would contact Infor support to obtain a DR license to run on the DR boxes, and then have to start all the sites, and contact connected systems to repoint to our DR boxes.  Not a small effort, to be sure, but not out of line for the destruction/disablement of an entire health system’s primary data center.

HTH,

TIM

Tim Pancost
Trinity Health

[Author]

Replies