Time between nodes in a backup schedule

[mosiac]

I've got a schedule that starts at 7pm for nodes that see very little activity that early in the evening. It's the only schedule that starts at 7pm and its just a simple incremental backup job that goes off every day.

What I noticed when reviewing the previous nights backup is that one node kicked off shortly after 7 and took a minute to backup changes. The next node in that set didn't start for an hour after that and also only took an hour.

My question is what should I look at to find what could be causing tivoli to wait an hour to work with 2 nodes in the same schedule? Could it be client side and the client was waiting for something to complete first?

Should I look at the event logs for the client that waited an hour to start, should I look at the activity log between 7 and 8 to see what other activities may have been occuring at the time?

Thanks

 

[marclant]

My question is what should I look at to find what could be causing tivoli to wait an hour to work with 2 nodes in the same schedule?

Perfectly normal, it's the randomization: http://www-01.ibm.com/support/knowl...rv.reference/r_cmd_randomize_set.html?lang=en

 

[mosiac]

That makes sense, but I'd like to try and make things a bit more efficient. Since we run backup schedules in the evening and all the daily tivoli tasks during the day waiting an hour for nodes to kick off seems a little wasteful. Would it be better to make a couple other 7pm schedules or lower the randomization time below 25 to say 20 or 15 to test?

 

[marclant]

That makes sense, but I'd like to try and make things a bit more efficient. Since we run backup schedules in the evening and all the daily tivoli tasks during the day waiting an hour for nodes to kick off seems a little wasteful.

If you can accomplish all the backups and housekeeping in a 24 hours period, nothing is wasted. It won't run more efficiently either, it will just run closer to the start time.

Would it be better to make a couple other 7pm schedules or lower the randomization time below 25 to say 20 or 15 to test?

For testing, just cut down the duration. The duration is not how long the schedule can run, but the window the schedule can start in. So if you have a 19:00 schedule with a duration of 10 minutes, they will all start in the first 5 minutes if you have a randomization of 50.

Wait until you have 500 nodes in a schedule, you will not want them to all start at 7, that's why there is a randomized start based on a percentage of the duration.

 

[mosiac]

Thanks for the info. I'll start looking at schedules with only 1 or 2 nodes and lowering their duration first