[v13] HPE Alletra Storage Server 4140

[chaos]

Hi everyone,
we're planning on using some Alletra 4140 for our new v13 environment as hardened repositories.

But there are a few questions where i haven't found an answer to it yet:


The Alletra 4140 is listed as a "Veeam Ready Appliance" but there is this remark noted

Important note: This solution has not passed the performance tests of Veeam Ready – Repository verification yet.

Veeam Ready Repository solutions are primary backup target solutions that have been qualified as meeting or exceeding both functional and performance tests for backup and restore operations.

Is there anywhere a definition what this requirements are and what "not passed" means - is it off by 10MB/s or more like 250MB/s?


The Alletra has two RAID Controllers for her 68 or 92 drives.
As to my understanding it does present it as two separate LUNs to the OS.

How does this work inside VHR/JeOS - will the XFS Volume be (LVM) striped over both LUNs?


Thanks a lot in advance.

[Steve-nIP]

I've used Alletra 4140 (stupid name, they should have stuck with Apollo, too many product lines mixed together under "Alletra") and it had one RAID controller for the 68 hard drives, presented as one giant volume

P61479-B21 HPE Alletra Storage Server 41X0 SR 32-port PCIe FIO Storage Controller

Looking at the specs, it only supports 32 drives directly attached, so there must be a SAS expander inside. QuickSpecs for the 4140 say you can one or two, with two, you're going to have two volumes, for sure.

[haslund]

Hi @chaos,
The exact requirements are not disclosed; however, in this case HPE simply did not perform the testing. The Alletra Storage Servers are widely used in the field though without issues.

I am confident they will test it once our testing tools have been updated for v13.

[Andreas Neufert]

Veeam Ready Appliance program moved from a vendor best practices program into a certification program, where all vendors have to pass first Veeam Ready Repository before they can become Veeam Ready Appliance.
The entry that the HPE systems did not pass Veeam Ready Repository does just reflect that the entry was done before the requirement and that the vendor still need to do the official testing with v13 later.

Veeam Ready Appliance stands now for minimum performance requirement (same as Veeam Ready Repository) + by vendor best practice documentation for the configuration.

The HPE 41x0 series is one of the most used repositories with excellent performance numbers usually way above 3GB/s backend (6GB/s backup speed) in many customer used configurations. In some cases 8GB/s backend speed with 16GB/s backup speed were achieved with the 92disk model. HPE can provide you with sample configurations and their throughput achieved with Veeam. As well you can find some guidance on our security blue prints here: https://www.veeam.com/security-blueprints.html

Regarding the "Apollos", you can get Raid controllers with 16 or 32 lines (which are individual disks basically) and so you need multiple Raid Controllers in the larger configurations.
You can create large volumes and use it in a SOBR as individual extents or you can create a LVM on it and give it to Veeam as one large volume, which is the most common configuration.

You can check on it in the HPE specific documentation: https://www.veeam.com/sys1394 (follow "vendors website" link there)

[chaos]

Okay thanks for the information.
So it's crucial to only have one RAID controller and by this only one LUN.

[Andreas Neufert]

No, I would say it is beneficial if you have multiple Raid controllers for the throughput.
It is just some extra configuration work to do.

[chaos]

Hi Andreas,
yes performance wise this is definitely beneficial.
But my understanding was, if I user VHR there can only be 2 LUNs be presented to the OS.
The first "small" one for the OS data and the second "bigger" LUN for the backup data.
With two RAID cards there should - as to my understanding - be 3 LUNs be presented (1x OS and 2x backup).
Is JeOS capable of "understanding" this?

[Steve-nIP]

Aren't you booting from the NVMe boot device that sits in the back of the server? That's a totally self-contained raid1 solution.
P48183-B21

[chaos]

Hi Steve,
we will boot from a NS204 - not quite sure without looking on the guides where it's located in an Alletra.

But looking from the OS side of it, there will be like /dev/sda (NS204), /dev/sdb (RAID controller 1) and /dev/sdc (RAID controller 2).
JeOS then uses the smallest available LUN for OS, so /dev/sda is taken.

Or I am completely wrong and the installer is somewhat more "intelligent"/configurable now to circumvent this.

[Andreas Neufert]

@chaos
Please see here:
https://helpcenter.veeam.com/docs/vbr/u ... tml?ver=13

If you have muliple disks, then disk 2,3 and so on will be added to an LVM and are used together as one volume and one repository.

[chaos]

Hi Andreas,
that sounds great.
Could you point me where this is exactly stated?
I just can't find it anywhere...

[Gostev]

Second occurrence of "Disk 2" on the page is in the sentence you're looking for.