Back in NetWorker 7.6 SP1, a new feature was introduced – Checkpoint backups. If you’ve not used or heard of then, you’ve probably not really even noticed how simple it is to turn on:
The notion of checkpoint restarts is easy – if for some reason the connection to a backup fails during the process, rather than restarting it from the very start of the backup, NetWorker can restart it from either the last directory the save process had entered, or, the last file it had started.
In theory, enabling checkpoint restarts is supposed to partially slow down the backup process – to keep track of checkpoints, NetWorker obviously has to do a little more work. In my tests, even with constructed filesystems having tens of millions of files, I’ve noticed very little performance difference between checkpoint restart backups and regular backups – and often in favour of checkpoint restart backups. That being said, there’s obviously a reason why EMC says they can slow the backup down, so your mileage may vary.
But here’s the thing – if you’ve still got large fileservers running on Windows, Unix, etc., then checkpoint restarts are a no brainer. A backup of 4TB fileserver that can either restart from 0TB if the backup fails 3.9TB through, or restart at 3.8TB if the backup fails 3.9TB through: which would you prefer to deal with?
Are you using checkpoint restart backups in NetWorker?
Have you calculated impact on index file sizing?
Hi Hrvoje – I’ve not observed a massive increase in index sizes as a result of checkpoint backups. I would expect in terms of current disk capacities it would be minimal, and entirely dependent on the size of directories that the client was in at the time of an interruption.
Preston,
I would like to start with telling you how much I like your blog. I have picked up some great tips here. Can you answer this question about “Checkpoint enabled” please?
What happens if you have a Networker group with some clients that have “Checkpoint enabled” and some clients that do not have “Checkpoint enabled”?
I have seen some issues with clients with “Checkpoint enabled” not restarting from the point of failure if other clients in the same group do not have “Checkpoint enabled”. This was tested on Networker 8.0.1. It may have been due to some unrelated issue but that is what it looked like on multiple test so I wanted to ask to see if you noticed anything like this in your testing.
Hi David,
Thanks for the feedback. Unfortunately I’m travelling at the moment and don’t have access to my lab, so I can’t test this.
I vaguely recall seeing some reference to keeping checkpoint clients and non-checkpoint clients isolated by group, but I don’t know where/when I saw this. Certainly a search of EMC Support isn’t turning up anything concrete for me.
It certainly seems from your experience that there may be a need for isolation however.
Cheers,
Preston.