Equallogic Direct San LUNs "online" issue.

[briguyiu]

Greetings,

Currently our Veeam backup server, and only proxy at this time, is physical. We would like to use direct access mode for transport, as NBD leaves a lot to be desired. I have done everything I have read to do to prepare for this, including the following:

1. Set automount disabled
2. Set automount scrub
3. san policy=OfflineAll
4. Rebooted.

Anytime I connect to a LUN that is also in my VMware environment, it comes up ONLINE no matter what. No drive letter is attached, and it doesn't show up in my computer, but I'm still very leery of adding a production LUN at this time. No HIT tools have been installed on the server yet, and I haven't configured anything differently for MPIO, than anything out of the box. If I manually set it to OFFLINE, remove it, and re-add it, it stays OFFLINE, but again, this scares me. Has anyone else experienced this issue? I'm leaning towards moving my Veam server, or at least a backup proxy to a VM to use hot add, if I can't get this straightened out. I don't see too many advantages to Direct San within a VM.

[Vitaliy S.]

Hello Brian,

Bringing LUNs to online state does not corrupt anything, but disk initialization does, so you should be fine.

Thank you!

[briguyiu]

Thank you for the quick response. So seeing Disk 4 online, but not showing up, up top, and having the below settings will not present any issues?

http://imgur.com/OwK0KNg

DISKPART> san

SAN Policy : Offline All

DISKPART> automount

Automatic mounting of new volumes disabled.

DISKPART>

[Vitaliy S.]

Correct, the issues will only happen if you decide to initialize these LUNs, which will lead to file system signature change and loss of data.

P.S. and it appears that you're not the only one this question > SAN mode backup. Disks offline or online?

[briguyiu]

I'm still leery. Without doing anything special to initialize the disk (I assume you have to do it via command line with automount disabled), I was able to delete the volume on the disk. In that other thread, it seems it is suggested that it DEFINITELY be off. I plan on doing some more tests, but if the speed of virtual appliance (hot add) vs. direct san is negligible, than I would prefer to move all proxy's/backup servers to a VM.

[briguyiu]

Another issue that we seem to be having during testing of this, is that with transport mode set to Automatic, SAN is never used for processing, NBD is. Once I manually set it to Direct Access, SAN is used. Then subsequently if I move it back to automatic, it will choose SAN properly. Is this a bug of some kind, or does it not evaluate which mode to use with each backup?

[briguyiu]

Ok, one last question. Testing a VM based backup proxy worked the best for us using hot add, but the bottleneck is the network. I assume this is because the backup repository resides on the physical server as an iscsi drive, and is locked down via ACL to just that server, so ESXi hypervisor stack can't access it. What is the best way to accomplish this? Opening up the ACL to all esxi hosts? If it was formatted via windows and NTFS, would the hypervisor still be able to access it? Does it have to be scanned in first? Thanks again for all the help!

[Vitaliy S.]

Then subsequently if I move it back to automatic, it will choose SAN properly. Is this a bug of some kind, or does it not evaluate which mode to use with each backup?

SAN mode should be used regardless of the proxy settings (Automatic/SAN). When you select SAN mode, do you have a checkbox that allows you to failover if network mode is not available?

Ok, one last question. Testing a VM based backup proxy worked the best for us using hot add, but the bottleneck is the network. I assume this is because the backup repository resides on the physical server as an iscsi drive, and is locked down via ACL to just that server, so ESXi hypervisor stack can't access it. What is the best way to accomplish this? Opening up the ACL to all esxi hosts? If it was formatted via windows and NTFS, would the hypervisor still be able to access it? Does it have to be scanned in first?

If you're using iSCSI initiated volume as a repository, then all the traffic goes from the proxy via network stack of the ESXi host to the repository. I'm not sure how ACL can help you in this case.

Anyway, having network as a bottleneck is absolutely fine, you will always have a bottleneck no matter what configuration you use. These stats just show you the "weakest" part of your configuration, that should be improved if you want to have higher job performance.

[briguyiu]

SAN mode should be used regardless of the proxy settings (Automatic/SAN). When you select SAN mode, do you have a checkbox that allows you to failover if network mode is not available?

Yes I do, and it's still checked. The issue is that without me explicitly choosing SAN mode, it uses NBD until I explicitly tell it to use SAN. That seems odd to me, as there is no issue using SAN mode, so I'm not sure why it defaulted to NBD.

[veremin]

Hi, Brian,

As long as Direct SAN proxy is specified correctly, it should be selected first, as per this sticky FAQ. So, the described behavior looks strange, indeed.

I would recommend contacting our support team and let them check might be the cause of proxy automatic selection failure.

Thanks.