Feature Request - create logic in copy jobs

[mrholm]

Hi
Been working with Veeam System Engineers and created a case (04842726) around our issue/problem and I'm a little bit surprised that this is a problem in a product like Veeam.
Since I'm not able to send in my documentation with all of mt findings I ask Veeam to look at the case to see my feature request. For rest of the audiance this is the short version:

For Backup
- We have 7 physical proxies
- Each proxy has its own local Repository, each with proxy affinity to itself
- One SOBR consisting of the above 7 repos

Backup jobs are set to use all proxies and this means that a backup is handled one proxy and stored on local disk, works great logs show:
- Using server16_repo01 scale-out repository extent
- Using backup proxy server16 for disk Hard disk 1 [nbdssl]


For Copy - we have the setup based on Veeam System Engineer recommendations
- Same 7 physical proxies
- One Data Domain as target for copies
- Each proxy has its "own" Repository in DD represented by own folder, each with proxy affinity to itself (I know this does not apply to copy jobs which is very strange)
- Each DD repo has gateway set to Automatic (we have also tried with setting same server/proxy)
- Each DD repo has its "own" server (proxy) as Mount server
- One SOBR consisting of the above 7 repos

So you could expect to see same behavior to use same server to read the backup file, perform DD-boost deduplication, and send to DD but not, logs shows:
- Using ddd001_emea01_server14 scale-out repository extent

So with handling +1000 servers daily you can imagine the amount of network traffic this causes when Veeam have one server to send backup data to second server to perform DD-boost and store on DD

I expect an Enterprise product to be able to get logic like this and this is also the feature request:

When starting a copy have Veeam look for logic:

What data are to be copied?
-data1 from server1
Where is the data1 backup files for server1 located?
- On proxy1
Where is the copy job writing its data?
- To DD SOBR1
Is proxy1 available as gateway server to any extent in DD SOBR1?
- Yes
Use proxy1 extent to read backup file and write copy

In an Enterprise envionemnt where backup and secondary offsite copes are demanded we need to have a backup application that can handle this kind of logic, I've been working with NetBackup, CommVault and Networker for 20 years and this logic has been there for ages so please make a good application as Veeam is even better and have some logic put into the copy jobs as well

//Mats

[HannesK]

Hello,
Note: backup copy jobs work "repository-to-repository" https://helpcenter.veeam.com/docs/backu ... ml?ver=110 Proxies are not involved in backup copy jobs. So proxy affinity is irrelevant.

I assume the reason for splitting the one DataDomain into 7 smaller chunks is the gateway server and how it works with backup copy jobs in automatic selection mode https://helpcenter.veeam.com/docs/backu ... ml?ver=110 - as you said, every DataDomain extent has its own mount server. The table shows the reason.

Improving the gateway server selection / scalability is a known feature request and we count your vote as +1.

It sounds like you want a 1:1 connection everywhere (instead of scaling out). Avoiding scale-out repositories could provide 1:1 connections (I assume that you will not try to change the design because it's a lot of work)

So my only question is: what kind of network connection do you have and how high is the load? I agree that it would be nice to avoid the cross-traffic. But for most customers it's not really an issue because the network connection of the proxy / repository servers is fast enough.

Best regards,
Hannes

[mrholm]

Hi and thanks for the reply
WHen we had only one repository Veeam always took the Mount Server set for the DD repo to be the gateway server, resulting in 7 front end proxies sent all backup copies to this mount server resulting in very high load on that server.
We don't want a 1:1 solution we want tp keep the Backup SOBR over all out 7 proxies to use all performance and same on copies, but we want Veeam to choose the correct server as gateway server based on where the backup file is stored. That makes most sense to me, as said we handle +600 servers with this setup and we send 3-4 TB of changed data into the DD after deduplication daily.
Our proxy servers are connected using 2x10Gbe to switch environment but the amount of data to handle is causing a lot of data sent, especially when we need to have most of the copy jobs to read from source since using synthetic fulls towards the DD ends up Veeam storing way to much data for servers with high daily change rate (+500GB on some servers). We seen situations where the running Full in syntehtic full are 60TB in size on a 560GB sized server due to the daily changed rate and synthetic fulls.

My feature request stays

We are on v10 and are planning to upgrade to v11 next week and I've seen a couple of notes about improved performace around backup copies so we cross our fingers on that. Also this will give us the possibility to do the read from source without having one extra restore point in GFS as now.

[mrholm]

Hi again
Really Veeam, please create logic in copu jobs to look what gateway server to use. Now with v11 when all copy jobs are done in fisrt or last week we see a hugh problem in setup using Data Domain as secondary repository and using several gateways to spread the load. A setup with many gateways in Veeam creates network traffic between the servers that together ith the cpu load form DD boost api makes the servers to be unreachable at times. We have to change all start times of backup jobs (backup proxy = DD gateways) to make all backups finish before any copy can take place.

So please add same excellet logic in copy jobs using several gateways in front of SOBR as there are in backup using proxies!

[HannesK]

first or last week

would second / third / fourth help? Because that's something we might be able to solve soon.