I have been rolling out VMware backups with Avamar for the past few months. One thing I have noticed is that sometimes the proxy transport used is not what we would like to see. There are 30 ESX clusters and around 5,000 VMs in this environment, about 20% have been moved over to Avamar. I have deployed a pair of proxies on each of the largest ESX clusters (by count of VMs). What I had expected based on previous conversations was that any VM that was on a cluster with a proxy would use the HotAdd transport for backups, and if the VM was on a different cluster nbd would be used.
For the most part that is the case. I am seeing unexpected behavior if the proxies on an ESX cluster are already fully in use. If a job order for a VM backup on that cluster needs to be processed, a proxy on another ESX cluster will backup that VM using nbd. I would prefer it if a proxy and a VM are on the same cluster the job would queue until the proxy had a stream available for that backup so it could use hotadd. I have been told there are options that can be set to control this behavior, but I can't find any documentation on this.
I found one tech note for older version of Avamar where you could make a change on the proxy to force it to do hotadd only, but that's not really what I want.
Anyone have an idea how I can get the behavior to work like this?
Always use a local proxy for the backup of a VM on a cluster where a proxy exists, but allow the proxy to use nbd for backups of a VM on a different cluster. Never backup a VM using nbd if there is a proxy on the cluster hosting that VM.
I am running Avamar 7.1.1-145 for both the server and the proxies with Data Domain for the storage behind it.
I have not done it but what I think you are looking for is in the proxy setup within Avamar
in Administration edit a proxy
you might have to work with this in setting up your proxies as to what Datastore it can back up.
Limit a proxy to datastores on one cluster, then an job that comes up from a different datastore cannot use that proxy.
or you could make groups based on where the client is and limit the proxy by the group.
you should have at least 2 proxy per - in case one proxy has an issue and stops working you would have a second one that would (hopefully) still work.
Try looking at this until someone else chimes in with a suggestion.
I see I can restrict a proxy to specific datastores, and in the end that may be what we need to do. I was hoping to avoid that because of the additional complexity that would add to the environment. At last count we had 600 datastores across 30 clusters, and the environment is very dynamic. Every couple weeks new datastores are added, and VMs migrated to them.
If you always want to perform hotadd backups you can do it by changing the following parameter in /usr/local/avamar/var/mc/server_data/prefs/mcserver.xml file
"proxy_selection_algorithm" to "hot_add_only"
Note : Shutdown MCS first by running command : "dpnctl stop mcs"
It should look like this after the change
admin@avamar:~/>: grep "proxy_selec" /usr/local/avamar/var/mc/server_data/prefs/mcserver.xml
<entry key="proxy_selection_algorithm" value="hot_add_only" />
Once the changes are made, start the MCS and scheduler
dpnctl start mcs
dpnctl start sched
Note : After making this change if there are no proxies available to perform hotadd backup it will eventually fail. Make sure you have sufficient proxies to protect your infrastructure