Questions about DrivePool as Veeam Backup Repository target...

[zim2323]

I'll pre-face this by saying that I think I already know the answers to this, but I want to ask to be sure.

Current environment consists of several direct attached storage volumes (all 15K SAS) and several fiber attached storage volumes with mixed 4-7TB 15k SAS volumes, and three 20TB 7200k SATA volumes.  In total there are 11 separate volumes with associated shares that are used for backup.

My thoughts were to implement DrivePool and combine all volumes into a single drivepool to create a "performance disk" that will have smaller retention sets (3 days) and then longer term "capacity disk" copied to our DataDomain (yuck!) for long term archiving.  We have a few backup sets that have primary full VBK files around 6TB with incrementals of around 100gb each, which could have several sets of those depending on retention policy.

1)  Is there a way to write files to the disk with the most available space in the pool?

2)  File sizes are unknown until the backup is finished.  My assumption is that this will be a problem for DrivePool, in that, if it's writing to a disk that only has 4TB free and it ends up being a 6TB backup, then it will fail.  Correct?

3)  I'm assuming there's no way to allow "write continuation" to another disk if the current disk fills or hits the % max.

4)  If a disk starts to fill can I set a lower max % , say 50%, and set the balance plugin to run every X minutes?  My intent would be that if a disk would start to "balance" other data off the disk and make room for additional write capacity as the current backup being written grows.

5)  I would anticipate that we'll use 70-80TB of the almost 100TB that we'll have available to us.  We will have headroom, but I'm concerned about managing/maximizing write space.  Depending on above answers.  I would assume Veeam will start having write failures for larger backup files if there's not enough room on the volumes.  

6)  Can I configure a non-SSD as a cache point, say one of the 20TB SATA volumes, that would then write out to the pool?  I'd used it purely as a staging point, rather than for performance.  At this point, ANYTHING is faster than our DataDomain's

Has Stablebit worked with any Enterprise customers on a JBOD solution like this that will help manage this type of scenario?  Or, is there a better SOFTWARE (not hardware) product that you would suggest that could do a better job?  We're probably another year out from purchasing proper backup storage so I'm stuck using what I have for now.

 

Thanks,
Chris

[Shane]

1)  Is there a way to write files to the disk with the most available space in the pool?

DrivePool defaults to attempting this, though if multiple large files start being written more or less simultaneously to the pool there might be an issue (q.v. next answer).

2)  File sizes are unknown until the backup is finished.  My assumption is that this will be a problem for DrivePool, in that, if it's writing to a disk that only has 4TB free and it ends up being a 6TB backup, then it will fail.  Correct?

Correct.

3)  I'm assuming there's no way to allow "write continuation" to another disk if the current disk fills or hits the % max.

Correct.

4)  If a disk starts to fill can I set a lower max % , say 50%, and set the balance plugin to run every X minutes?  My intent would be that if a disk would start to "balance" other data off the disk and make room for additional write capacity as the current backup being written grows.

You can set Balancing to always run immediately upon detecting a trigger condition or only no more often than as small as any integer multiple (including 1x) of 10 minute intervals. Note that (I believe) it cannot balance open files, or at least not files that are actively being written. @Christopher (Drashna)?

5)  I would anticipate that we'll use 70-80TB of the almost 100TB that we'll have available to us.  We will have headroom, but I'm concerned about managing/maximizing write space.  Depending on above answers.  I would assume Veeam will start having write failures for larger backup files if there's not enough room on the volumes.

Correct. I've had this happen. I take it the enterprise version of Veeam still doesn't support splitting? (I use the standalone agents at home)

6)  Can I configure a non-SSD as a cache point, say one of the 20TB SATA volumes, that would then write out to the pool?  I'd used it purely as a staging point, rather than for performance.  At this point, ANYTHING is faster than our DataDomain's

Yes, you can. The SSD Optimizer plugin doesn't actually care whether an "SSD" is actually an SSD or not; it would be more accurate to call it the Cache Optimizer plugin.

For example, you might set "Incoming files are cached on drives A and B; when A and B are more than 25% full they empty* to drives C, D, E and F in that preferred order, but try not to fill any storage drive to more than 90% capacity and if any are then move files off them until they are no more than 80% full".

Note that you can also make pools of pools (so pool P could consist of pools Q and R which could consist of drives A+C+D and B+E+F respectively) if for some reason you want to have different configurations for different sub-pools.

*the SSD Optimizer plugin doesn't have fine control over emptying; when it starts it will attempt to continue until the cache is empty of all files not being written to it.

P.S. it is possible to write your own balancing plugins if you've got the programming chops.

P.P.S. do not enable Read Striping in DrivePool's Performance options (it defaults to off) in production until you have confirmed that the software you use works reliably with it. I've found some hashing utilities (for doing comparison/integrity/parity checks) seem to expect a single physical disk and intermittently give false readings when read striping is enabled.

[zim2323]

Thank you!

Yes, Veeam is still stuck 10 years ago and can't split.  LOL

I've had the same issues with Read Striping as well.  My issues came when I was read striping on SATA mechanical disks that were hosting VM's.  The speed, or lack there of, on the mechanical disks wasn't enough to account for some of the additional overhead required for read striping.   At least that's what I theorized.  After I moved to SSD, I had no issues read striping.

 

Thanks again!

[Shane]

On 6/1/2023 at 1:55 PM, Shane said:

3)  I'm assuming there's no way to allow "write continuation" to another disk if the current disk fills or hits the % max.

Correct.

I should perhaps add here that DrivePool won't stop writing a new file if it hits the % max, only if/when it hits actual full. The % max setting is for files yet to be written (e.g. "this disk is at/over % max, don't put more files here") and/or files that are not in use (e.g. "this disk is at/over % max, move file(s) off it until satisfied"); an example of a balancer controlling both of these (no new files and move old files) can be found in the Prevent Drive Overfill balancer.

On 6/1/2023 at 1:55 PM, Shane said:

Yes, you can. The SSD Optimizer plugin doesn't actually care whether an "SSD" is actually an SSD or not; it would be more accurate to call it the Cache Optimizer plugin.

For example, you might set "Incoming files are cached on drives A and B; when A and B are more than 25% full they empty* to drives C, D, E and F in that preferred order, but try not to fill any storage drive to more than 90% capacity and if any are then move files off them until they are no more than 80% full".

Likewise, with the SSD Optimizer plugin its option "Fill SSD drives up to X%" doesn't mean "empty at this point", it means "overflow to Storage drives at this point".

When it empties is instead controlled by your choices in the main DrivePool settings menu (Balancing -> Settings -> Automatic balancing - Triggers).

This may be somewhat counterintuitive, but it is useful to avoid running out of space in situations where the "SSD" is being filled faster than it can be emptied.