Guest Interaction Proxy Selection Process?

[rreed]

Afternoon all. How does v9 go about selecting a given job's Guest Interaction Proxy? When we set ours to Automatic, one data center's job might pick a proxy from another data center half way across the country rather than internal to its own data center. I've had to go through and set all jobs to manual, albeit all the local proxies per each site, but it seems the general Proxy selection is smart enough to pick from the local building when set to Automatic, but doesn't seem the Guest Interaction Proxy is.

[LinusH]

Hi,

I have the same problem (almost anyway, Veeam uses a proxy in another city) and tried to create a case to see what Veeam said (Case # 01683064). Their response was as quick and friendly as ever but for once I didn't like the answer:

When choosing a log shipping server for the job, Veeam Backup & Replication considers the location of the Microsoft SQL Server VM and log shipping server. Veeam Backup & Replication uses the following priority rules to select the log shipping server:
1.Log shipping server is located on the same ESX(i) host as the Microsoft SQL Server VM.
2.Log shipping server and Microsoft SQL Server VM are located in the same network.
3.Log shipping server and Microsoft SQL Server VM are located in different networks (the production infrastructure is isolated from the backup infrastructure).
That is, when choosing a log shipping server, Veeam Backup & Replication will give the top priority to a Microsoft Windows VM that is located on the same ESX(i) host as the Microsoft SQL Server VM and that has a network connection to the Microsoft SQL Server VM

I suggested a few additions to the rules and the reply was: Sorry, perhaps you can check the forum?
So, here we go, this is how my problem would be solved (rule 3 and 4 are only in my imagination). What do you think:

1.Log shipping server is located on the same ESX(i) host as the Microsoft SQL Server VM.
2.Log shipping server and Microsoft SQL Server VM are located in the same network.
3.Log shipping server is located on the same vmware cluster as the Microsoft SQL Server VM.
4.Log shipping server is located on the same vmware datacenter as the Microsoft SQL Server VM.
5.Log shipping server and Microsoft SQL Server VM are located in different networks (the production infrastructure is isolated from the backup infrastructure).

[rreed]

Your suggested rules would be great for guest interaction proxy selection, Linus. In our case here we do have separate networks for ESXi management, vmotion, VM data (a few networks for those, actually), and backup; repeat for each data center w/ each being unique entirely. No two subnets are the same for either local data center or among data centers. Going by VMware cluster first then data center would really streamline and simplify this!

[foggy]

First of all, the rules outlined above are for selecting Log Shipping Server for processing of SQL Server transaction logs. Those for selecting Guest Interaction Proxy do not contain the first point.

That said, we'd like to investigate the behavior closer, so opening a case is recommended. The reason could be either network configuration issue or a bug in subnet detection algorithm, logs will tell whether the check was performed and why the particular server was selected.

[LinusH]

That said, we'd like to investigate the behavior closer, so opening a case is recommended.

Thank you for your reply, foggy. As you can see in my post, I had a case but that is now closed since I have manually selected guest proxy as a workaround. Actually it was not only related to this problem so someone should open a new one anyway.

[foggy]

Or you can do once/if you see similar behavior again.

[rreed]

Seems like it was every time w/o fail, it would pick a guest interaction from our other data center if set to auto. So weird. I've manually assigned for right now and dealing w/ other fires, will report back when I get back to this one.

[rreed]

One thing I certainly have noticed, is that despite having disabled a couple of our proxies, those disabled ones still get selected for guest interaction. I would think when you disable a proxy, it should be disabled for being selected for anything. Veeam?

[Gostev]

I have the same problem

Well, it is most definitely NOT the same problem as Guest Interaction Proxy (what OP is talking about) and Log Shipping Server for SQL backups (what you are talking about) are two absolutely different roles which have nothing in common, so discussing them interchangeably make this thread super confusing. But this far into the discussion I am not even sure how to clean up this topic to make it usable for future readers.

Ideas are good though.

[foggy]

One thing I certainly have noticed, is that despite having disabled a couple of our proxies, those disabled ones still get selected for guest interaction. I would think when you disable a proxy, it should be disabled for being selected for anything. Veeam?

This is expected, any Windows server added to Veeam B&R console can be selected as guest interaction proxy, despite the fact it is used as a proxy server (and whether it is disabled or not).

[rreed]

Thanks foggy. Any chance we'll have a future fix for "just disable this server, period?" I know it's a bit of a whine but it might be handy to have a one-stop disable to pull out servers for maintenance, trouble shooting, etc. rather than have to pull it out of rotation in several places.

[foggy]

So something like a "maintenance mode" for the infrastructure server, I suppose.

[rreed]

Actually, that would be a terrific enterprise-level feature of the product - "maintenance mode" at the managed server level, that way it takes it out of everything in one spot. Awesome.

[LinusH]

Well, it is most definitely NOT the same problem

Um, yes I think I have had the same problem. But I have also had problems with Log shippings after the upgrade to v9 for two jobs (the primary problemin the case) and the solution was to delete the jobs and recreate them. Anyway, this is how it looks like when my job uses the far-far-away server, since it is on a limited network:

Code: Select all

2016-01-30 23:14:12 :: Using guest interaction proxy farfaraway.otherdomain.local (Different subnet) 
2016-01-30 23:14:20 :: Failed to inventory guest system:  Error code: 0x800706ba
Failed to invoke func [GetHostInformation]: The RPC server is unavailable.

This is why I have removed it and would like to see those rules implemented. But it's not critical anymore since you nice people gave us users the little function "Read the entire restore point..." for our backup-copy job. With this I don't think it will read much from the far-far-away storage and thus I don't need a gateway server there. Now I'm back at "it just works" again.

I would anyway really like to see a maintenance mode for any server, it would be great! Smart suggestion!

Br Linus

[foggy]

But it's not critical anymore since you nice people gave us users the little function "Read the entire restore point..." for our backup-copy job. With this I don't think it will read much from the far-far-away storage and thus I don't need a gateway server there.

I'm not sure how this option can help here, though.

[DeadEyedJacks]

Just to add we are also seeing issues with the automatic selection algorithm for Veeam roles picking offsite servers.
Presumably the algorithm used is the same regardless of the particular role?
Limiting auto selection to within the same VMware datacenter would seem to be a sensible option.

[ChuckS42]

[Necropost resurrection engaged]

Was this ever really answered? I recently found a particular backup that insists on grabbing a Guest Interaction Proxy from a different site. I leave the setting on Automatic in general and don't have issues, but for one server.

So, what is the criteria for selecting a Guest Interaction Proxy? (Since the above has been muddled by the statements about log shipping server selection.)

[foggy]

Hi Chuk, in general, the proxy selection process is outlined in this user guide section. Have you contacted our support team for assistance in identifying the reasons of why the remote server is being selected? Any chance no local server has access to the affected VM?

[ChuckS42]

It must be hitting the third layer - machine in other network not performing a backup role. Unfortunately this means our other datacenter. (The first and second options would never be hit, since our backup servers are in a dedicated subnet/VLAN.).

The only way around this is to specifically define the GIP. Quirk of our network design, it seems.

[foggy]

This is exactly why the manual selection is available - for the cases where automatic isn't applicable.