RAM Maxed out

[andyboy]

I'm testing different options for Storage optimization. In my test, I have one job running, that has one VM in it (vm size is about 1.5TB) using the LAN target optimization and the RAM on the backup server is maxed out. I have 10gigs of RAM for the backup server. Is this normal??

[andersonts]

I think some more information is needed here. Is the screenshot showing the Veeam server and if so is this a default install? How much memory is being consumed by SQL (assuming the default install) versus the Veeam engine? Is the repository on the same server as the Veeam server and if so what kind of backup job are you running (initial full, reverse incremental, etc.)? I certainly would not expect a single job to use 10GB of memory unless there is something else going on like significant disk latency. What does the bottleneck show in the job when it's running?

[Yuki]

You know i'm seeing the exact same issue and so far have not found the culprit! I've had to raise RAM allocation to 16GB from 8GB as running backups caused machine to used up all physical memory. Once i've increased to 16GB, the usage seems to peak around 10GB. The truly strange part is that Windows doesn't show which process is using up all that RAM in task manager or resource monitor. Furthermore, once the backup completes - the RAM usage does not go down as if there was a memory leak. I've tried killing all SQL and Veeam processes but the RAM was not freed up, so i had to reboot the proxy to return it back to normal.


Our proxy on the WAN target:
OS: Win 7 Pro 64bit
CPU: 8 vCPU
RAM: 16GB
Hypervisor: vSphere 5.0
Veeam: 6.5 with latest patches
datastore: CIFS on NAS and Local
backup size: 7TB raw, 5TB backed up
type: reverse incremental
settings: they are different for different jobs (optimal or extreme compression and targets are either LAN or WAN).

[tsightler]

Most likely the memory is being allocated to the Windows file cache. This memory is reclaimable if there is actual memory pressure from processes. You should be able to use tools like RAMmap and VMmap to see where the memory is being allocated.

Can't recommend this video enough to understand how Windows keeps closed files "pinned" in memory after use.

[andyboy]

One job running a reverse incremental. It's not running a full backup. The SQL process was using about 4GB and Veeam was using about 1.5GB. Default install. The repository is on the same server using iSCSI (Twinstrata Cloudarray). Just Veeam is being used on this server, nothing else. When the job was done, memory usage went back down to 1.5GB.

[tsightler]

I'd say that's probably normal. Between the backup file being mapped in memory, and the space used by the Veeam processes, and the Windows write cache I'd expect this.