We are experiencing a very high CPU usage due to many (in our case >100) simultaneously starting snapmirror transfers at the same time (for example at 5 minutes after every full hour, if using the cron “hourly”).
I’ve written a script with distributes all snapmirror relations across four schedules using the NetApp Perl SDK.
The cron schedules have to look like this:
cluster::> cron show hourly.* hourly.0 @:00 hourly.15 @:15 hourly.30 @:30 hourly.45 @:45
Now you can run my script which distributes all relationships round robin over all four schedules.
You can find my script here: GitHub
After modifying all schedules, our CPU usage went down massively 🙂