03-05-2014 10:06 PM
So I can't seem to find any best practice information on V<ware resource limits in NetBackup 7.6. What experience have had with setting them? We are looking at virtualiasing 100% of our environment, using aVNX5700 array, on UCS blades with ESX 5.1, backing up to a 5220 appliance over SAN using the query builder. How do we monitor what effect the changes will have?
I have heard limit datastore to 4 backups is one, but that's about it.
Thanks.
Solved! Go to Solution.
03-06-2014 01:11 AM
The limit of how many per datastore you do at any one time is probably the most important - there was a tech note on VMware somewhere that actually says 1 is the best value.
The other thing is not to overload the appliance itself as it has to handle all of the snapshots - so if you have 50 datastores that would mean running 50 jobs at the same time - so limit the jobs per vCenter down to something like 20.
It does mean a round robin of backups but they will actually end up faster overall.
As you are looking at 7.6 remember that you can also use accelerator once your esx estate is at a suitable level - and a lot of those jobs running at the same time would put a big load on the appliance so make sure your /disk/etc/puredisk/contentrouter.cfg has the WorkerThreads section with a value of at least 128.
So from what i have seen 1 per datastore is best and maybe start at 20 in total and then increase that value until you find the optimum for your system
03-06-2014 01:11 AM
The limit of how many per datastore you do at any one time is probably the most important - there was a tech note on VMware somewhere that actually says 1 is the best value.
The other thing is not to overload the appliance itself as it has to handle all of the snapshots - so if you have 50 datastores that would mean running 50 jobs at the same time - so limit the jobs per vCenter down to something like 20.
It does mean a round robin of backups but they will actually end up faster overall.
As you are looking at 7.6 remember that you can also use accelerator once your esx estate is at a suitable level - and a lot of those jobs running at the same time would put a big load on the appliance so make sure your /disk/etc/puredisk/contentrouter.cfg has the WorkerThreads section with a value of at least 128.
So from what i have seen 1 per datastore is best and maybe start at 20 in total and then increase that value until you find the optimum for your system
03-23-2014 03:01 PM
Thanks Mark, (and sorry for late reply). This is really useful information. To measure the impact of these settings, is it best to look at how long backups take, or are there memory/cpu/network counters on the appliance we should look at? We can look at datastore metrics from the SAN array and throughput from within VMware. The actual ESXi host doesn't take much of a hit (apart from network) does it (RAM/CPU)?
03-25-2014 04:07 AM
Lots of things you could look at but the aim is top get the minimum backup window so i would concentrate on just that to be honest