Copy Back Job - Using mount server set in repository

[adb98]

Ran into an odd issue yesterday in which I had a Backup Copy job failing and stating it could not connect to the storage. Long story short, I had changed the mount server on the repository that I was pulling the backups from to a 2016 physical server I had built specifically so my REFS partitions would mount for FLRs. This was only supposed to be a mount server so I never hooked it into my backup network.

I changed the mount point on the backup repository back to a server that had access to the repository and this resolved the issue. With identifying the cause I went ahead and added the mount server to the backup network so it could see the repositories.

My question is .... shouldn't the GFS be like any other job and use the gateway to determine which proxy it will use to pull the backup out of the repository. Seemed kinda odd to me and didn't make sense. Also I am sure others may run into this if they tried to get around the whole REFS 3.1 / FLR mount issue when using 2012r2 backup servers.

[Andreas Neufert]

Hello Aaron,
thanks for the question.
When you have Windows/Linux repositories, the BCJ should just use the Repository Server itself.
When you use a CIFS share with dedicated Gateway Server, the dedicated Gateway Server will be used for transport.
If you have a CIFS share with "Automatic" selection of Gateway Server the mount server is used for Backup Copy Jobs and Merge processes with fallback to Backup & Replication Server.

In some situations this is not perfect and we had discussed some changes there.

[adb98]

Thanks Andreas. Glad to know I was not crazy and that really helps. That is a little ugly and I would say needs clarification in the docs and I agree it needs some work. I take it that if its a Data Domain appliance is in the same catagory as CIFS as that is what it is doing.

So the fix for me would to take my prod repository and my GFS repository that have a 2016 REFS VM in them and set the mount option to the mount server (2016) and the the gateway hard set to one of my proxies. This way my GFS will still use one of the main proxies instead of a mount server.

[adb98]

Andreas is there a way to PM you? Have something I would like to share but would like to keep it out of the forum. Don't see an option to direct message you.

[Andreas Neufert]

Email: andreas.neufert (a-t) veeam.com

[pirx]

Hello Aaron,
thanks for the question.
When you have Windows/Linux repositories, the BCJ should just use the Repository Server itself.
When you use a CIFS share with dedicated Gateway Server, the dedicated Gateway Server will be used for transport.
If you have a CIFS share with "Automatic" selection of Gateway Server the mount server is used for Backup Copy Jobs and Merge processes with fallback to Backup & Replication Server.

In some situations this is not perfect and we had discussed some changes there.

As we talked about that a few weeks ago, I've some open questions.

- if I configure a gateway server in the repository this one is used for reading data from a CIFS share (NetApp)
- if I configure "Automatic Selection" the mountserver is used for reading data from a CIFS share (NetApp)
- as I can only configure one gateway or mount server this is a fixed configuration for a repository, even if "Automatic Selection" was chosen?

- for copy jobs with a different target CIFS repository there might be a different gw server configured
--> so 2 gw servers are involved, one for reading data, the other one for writing data. Is that assumption correct?

- I've to configure a dedicated gateway for S3 capacity tier. So it's the same here, source and target repository might use a different gateway server and there is no easy way to configure it to use the same gateway?

We have a lot of SOBR extents and 6 physical gateway/mount/proxy server which are shared by all extents, so it's usually not possible to use the same server for both (source/target extent). Are there any plans to change this behavior? In our case it would be good if the same server would be used for reading and writing/offloading data.

[Andreas Neufert]

There is a table here that explains it:
https://helpcenter.veeam.com/docs/backu ... ml?ver=100

Regarding S3 Capacity Tier.
When you do not select a Object Storage Gateway, each defined Gateway is used to offload data to object storage. In case of auto SMB/NFS gateway selection, the mount server (fallback to B&R) is used then.
If you have selected an Object Storage Gateway server then all mount servers (see one line above) will work through this object storage gateway server to communicate with the object storage. Think about it like an Internet Proxy when the other servers do not have access to the internet object storage. You do not need a object storage gateway server when you can communicate directly with your object storage!

One of the things we want to look into in one of the next versions is to give you the option to select multiple SMB/NFS gateway servers for load balancing and HA purpose.
When you look at the dataflow above you will see that the dataflow is already optimized for things like backup copy jobs. As well for object storage offload it is optimized.

In case of BCJ processing we choose the defined systems which will allow you to scale.

[pirx]

So if all configured gateway/mounts servers of the performance extent have internet access, I don't need to configure a gw server on S3 capacity extent?

[Andreas Neufert]

Correct. If they can communicate with the object storage directly there is no need for this. Remember that DNS need to work as well then.

[pirx]

"When you do not select a Object Storage Gateway, each defined Gateway is used to offload data to object storage. In case of auto SMB/NFS gateway selection, the mount server (fallback to B&R) is used then."

Just to be clear, that does mean there is only one gateway involved for reading from CIFS and offloading, the one that is configured as gw/mount server for the extent the data is read? Or is it that there are still 2 involved and the one for offloading is some random one from list of all possible gateways? Maybe I'm mixing up things but I have to say that the dataflow is not completely clear to me.

[Andreas Neufert]

SMB to Object Storage
SMB Gateway fixed
Object Storage no gateway
SMB=>SMB Gateway=>object storage

SMB to Object Storage
SMB Gateway Auto
Object Storage no gateway
SMB=>Mount (fallback B&R server) =>object storage


SMB to Object Storage
SMB GW fixed
Object Storage Gateway selected
SMB=>SMB GW=>object storage gateway => object storage

SMB to Object Storage
SMB Gateway Auto
Object Storage Gateway selected
SMB=>Mount (fallback B&R server) =>object storage gateway => object storage

[pirx]

Thx! One last clarification regarding BCJ.

"In case of BCJ processing we choose the defined systems which will allow you to scale."

There are always both gateways involved, the gw/mount server configured on source repo and the one on target side?

[Andreas Neufert]

Each of them work with the local defined GW (or if Auto Mount) server.

BCJ with SMB Gateway selected
SMB1 => GW1 => GW2 => SMB2

BCJ with SMB Gateway auto
SMB1 => Mount1 (fallback B&R Server) => Mount 2 (fallback B&R Server) => SMB2

If you want to process this on the same server then you need to select the same mount server.

It does not make sense to use a single server here by default, as you would run into network bandwidth issues as there is very likely a smaller link involved between sites.

[pirx]

Understood.

As we have multiple SOBR's each with a couple of extents and we distribute the 6 mount/gw servers as good as possible among the extents, I feel that sending the data from a mount server of the source repository to a different gw server on destination side is not optimal. I also don't see how we could create a 1:1 relationship where source and destination server is always the same. With the different SOBR extents we can not control on which extent the data goes on destination side.

What do I miss?

[HannesK]

Hello,

What do I miss?

as I remember your setup vague from our last meeting... I talked to your Veeam SE and it looks like that there is a partner meeting / health check planned... would it be possible that you draw a picture with all the gateways involved for that health check meeting?

Because it's really complicated to follow the questions in a forum with that level of complexity of your environment.

Thank you,
Hannes

[pirx]

I can do this. As the health check will be performed by a Veeam partner I can't tell when this will happen.