Stackato 2.10.4 health manager is aggressively autoscaling up DEAs

Posted by lorned on 2013-05-09 12:34
Question: 

My health manager is too aggressive about how many DEAs it scales up after I have had an outage to some/all of my applications. What can I do to make health_manager autoscale less DEAs?

Answer: 

Health_manager uses an algorithm to determine what its expected resource usage is in the future. Part of that algorithm is how much memory is being used by the system currently, and how much that value has changed recently. Following an outage or some other service interruption to one/all of your DEAs, as all of your applications begin to start at the same time, the algorithm can be too sensitive to the sudden change in resource demand and can attempt to auto scale too many DEAs at the same time, which won't really be necessary long-term.

We have patched this algorithm to reduce how quickly it reacts to these kinds of sudden changes. The patch can be downloaded from http://get.stackato.com/patch/2.10/stackato-2.10.4-hm-autoscaling.sh. This patch will need to be applied to your primary node (the only node you have running health manager). You can apply it by running 'sh stackato-2.10.4-hm-autoscaling.sh' from a console/terminal session. After applying the patch you will need to restart the health_manager process via 'kato process restart health_manager'. This will not cause any important system outages and should not be visible to any users.