Accepted Solution

6290 Controller utilization and failover


I have a 6290 that shows the "any" cpu time of about 50 to 80 % over a 24 hour period.  The per cpu utilization is about 25 to 35 % over a 24 hour period.


Which measure should I use to determine when the controller is overloaded and will not do a failover to the surviving controller is needed?


If I go by the "any" cpu measure then I would think 50%. ( Fail over 50% onto another 50% and you are at 100% =slow)


I think part of the issue here is that I am confused about when to use the "any" cpu measure and when to use the per cpu measure.


sysstat -m



Re: 6290 Controller utilization and failover

At the risk of not answering your question, I'd say you need to look at more than CPU % busy to determine how loaded the filer is. How busy are the disks, what about the network? The CPU is really just one of numerous potential bottlenecks. What really matters is the latency from the client side (assuming you're serving some sort of client workload here).


The other issue with looking at CPU % busy is parts of OnTap can make use of more than one CPU.



Re: 6290 Controller utilization and failover

[ Edited ]

Good question and this comes up every so often with the "any" counter.  I wish the any cpu counter wasn't there... is any cpu busy at any percent.  sysstat -m and -M are useful to see per cpu core and domain stats... for a single value, average cpu is more likely what you are looking for.