Nervous About Direct SAN

[Gilmour]

I am interested in using direct SAN but I am a little nervous about presenting the VMFS on our SAN to a Windows host. I would like to hear from others who have done this and any words of wisdom you might have. I have ensured the SAN policy is set to "Offline All" and that Automount is set to "Automatic mounting of new volumes disabled". According to Dell, our SAN does not have the ability to provide read only access to the Veeam server.

Our configuration:

Windows 2012 physical server running Veeam
Dell Powervault MD3220i SAN

[Vitaliy S.]

Hi David,

Please take a look at existing topics with similar questions > http://forums.veeam.com/search.php?st=0 ... +automount

On top of what you've already done, I would suggest ensuring that only "right" people belong to the local admin group of the proxy server.

Thanks!

[Gostev]

Hi, David

Also, I cover this topic in a little more details in the VMware Backup with Veeam Best Practices: 2014 Edition VeeamON breakout session, please check it out. This has the latest and greatest information

Thanks!

[adrien.herve]

Hi David,

Don't be afraid, SAN direct mode is the best.
You just need to be sure of few things:
- You have the proper MPIO driver (if you don't want to see a number of volumes multiplied by the number of paths).
- Every volumes needs to be offline in the disk management.
- Check with DISKPART> san that your SAN Policy is Offline Shared.

Tell me if I'm wrong but how can you restore VMs on production datastores in SAN mode if you're not presenting RW volumes?

[Gilmour]

Everyone - thank you for the encouragement and useful information. I plan to implement this next week and I will up date you on the progress.

Adrien - in response to your question - I am in agreement, I don't see a way to perform restores without having RW access to the volume. However, I have read articles/posts in other forums that suggest that the SAN should be configured as RO if possible.

Thanks again everyone!

[dellock6]

Correct, but note that all those reccomendations were made before Veeam B&R v8, where we first introduced SAN restore. Before, it was ok to further increase datastore protection with RO, since there was no option to write back into it. Now you can set it to RW and rely on SanPolicy in Offline mode, so you do not have to reenable RW for a restore when needed.

[Gilmour]

I finally had time to configure direct san and everything went smoothly - thanks everyone for the words of wisdom!

However, this brings up a follow up question for the group.

I am seeing improved speeds over the network backup; however, I really thought I would see better speeds. On average, I am seeing between 100MB/s and 120MB/s. I realize that backup performance is dependent on several factors but I would like to hear if this is in the ballpark or if I should expect better performance.

[adrien.herve]

I think your SAN is maybe a little bit slow (but as you said previously it depends on several factors). Can you please create a test job with a large VM and start an active full? With this test you'll be able to reach and see the maximum speed of your source.

[foggy]

And bottleneck stats for this job will tell whether the SAN is the actual culprit or bottleneck is somewhere else in the processing chain.

[Gilmour]

I ran a full backup prior to start of business this morning. The backup was on a 80GB VM and below are the results:
Processing Rate - 71 MB/s
Bottleneck - Source
Busy: Source 73% > Proxy 13% > Network 40% > Target 39%

[Vitaliy S.]

Seems like you've reached the max performance of your source datastore. You can also try to use hotadd proxy, compare the numbers and then stick to transport mode with best performance rates. Thanks!

[dellock6]

I don't know the performances of aPowervault like yours, but it could be indeed reached its maximum.
Another possible cause, since it's a "i" model so iscsi if I remember correctly, can be the ethernet connection is the bottleneck, and is marked as source. Are you using 1Gb connections? If yes, you can try to enable mpio on the windows 2012 proxy and see if performances are increasing.

[Gilmour]

Dellock6 - You are correct, the "I" model is an iscsi and it is 1Gb. I thought about enabling mpio on the windows 2012 proxy but I have read mixed opinions on this. I will try enabling mpio next week and let you know how it goes - thank for the feedback.

[JosueM]

Good day everyone,

Anyone has done direct SAN with these disk raids? I'm a bit nervous about the iscsi connection from the windows proxy machine could damage/corrupt the VMFS datastore, since the disk cabin has no way to set read only access.

Any help is welcome.

Thanks in advance.

[foggy]

Josue, there is no need to be nervous, since Veeam B&R server automatically sets SAN policy to offline mode/disables automount during installation of the proxy server, which prevents disks from being initialized (it is disk initialization that can result in data corruption).

[JosueM]

Thanks for reply foggy, finally the connection was fine I did backup job but performance was extremely poor than hotadd mode. I opened the ticket 01091515 so I hope to get some light from the support especialist cuz I checked the config and all seems to be fine but performance is not as expected.

[PTide]

Hi,

performance was extremely poor than hotadd mode.

Could you please make sure that it did not perform a failover to a network mode, you need to check session logs for that.

Thank you.

[JosueM]

Hi,

The job log shows it used SAN mode

[foggy]

According to the screenshot, network is the bottleneck, not the reading speed from the source storage. So pay attention at the connection between proxy and repository.

[kevin@wei.com]

I love using the SAN mode myself. We have 3pars, so the storage integration really helps as you don't have to present the base datastores to the backup server. 3par and veeam take care of the work. Plus, you are only presenting a snapshot. If someone gets access to your backup server and sees all the 2TB datastores that could be used for something, if they format a drive, it is just a snapshot and gets destroyed at the end of the job anyways. No harm no foul. You may lose a restore point, but a lot less trouble than losing an active datastore. Netapp is supported and EMC will be in the next release (v9).
My only suggestion would be if you are presenting datastores to your backup server, make sure everyone that has access to the machine understands why they are there and how to handle them. In addition, you'll need to present any new datastores to the backup server. Otherwise your backups will default to NBD.

[JosueM]

ya it seems the problem to be the network, I did change the network adapter, previuos was broadcom 5720 now I installed Intel Gigabit card and the behavior is the same. I'm suspecting some setting in the Windows ISCI initiator, the setup also uses jumbo frame end to end and I did enabled it on the veeam backup server NICs.

I discard the switches cuz the exsi host are connected to the san in the same switches and performance its acceptable . support still researching and I hope to find something.

Thanks all for your help.

[t.gebler]

Related question:

I ve setup SAN Mode at a customers (Server 2008R2, FC SAN) years ago (About V6.5, now 8.x), dissabled automount manually back in the day. Works like a charm!

Now we would like to stage the local Backups to rotating USB disks, with Automount dissabled the disk wont show up ofcourse. Am I missing something? Can I enable Automount? Is there a difference to other Automount options policys? The customers SAN has no Read Only option, so thats not an option to be save.

Thanks four your advice

[ncolt]

I found this article useful http://www.danilochiavari.com/2014/01/2 ... e-vsphere/
I also sent the domain admins a warning email with a screenshot of Disk Manager just so there are no misunderstandings

[foggy]

Right, the article describes it correctly. In later versions (starting from Windows Server 2008), SAN Policy is configured, so there's no need in disabling automount (like it was required in earlier versions).

[notesguru99]

Hi,
We have been using Direct SAN Access from the word go with Veeam approx 12 months. I was a little apprehensive too, presenting my NetApp LUNs to my Windows box but I've had no issues and the Windows box reboots for updates no problems and I've even presented LUNs from other SANs and its not been an issue. I just do a sense check, counting the connected LUNs and the size of the disk whenever I'm in disk management.
It seems to me that you would have to consciously make the effort to destroy your LUNs via this method.
All the best
Stuart

[jmv]

Hello
Some time ago, during an installation, our Windows server got all the vmware LUNs and firmed it.
We had configured automount and SAN policy, I don´t know what was the problem, but that happend.
I suspect it was during a reboot after Windows OS patches installation.
These Luns have no data, anyway from that time I never use direct SAN if the storage can´t use read-only presentations.
Using several distributed proxies have a very nice speed, it´s a valid alternative.

[mpasaa]

I used to connect my physical HP Blade Backup server to the SAN and never had issues BUT I understand your concerns. I, too, wondered if Windows would decide to initialize unknown disks and such but this worry was unfounded. The volumes were mounted and displayed as UNKNOWN as far as Windows was concerned BUT because Veeam has the intelligence it has as long as those disks were connected Veeam was able to access the Veeam backup data and Windows was none the wiser. Of course, I would make sure NO ONE uses Computer management disk manager just to be safe but you can also run a DISKPART command I believe to prevent auto-mounting of disks which is what I did. Never had issues and performance was definitely very good. That said, why not use a Veeam VM and attach an RDM? I use that configuration now and performance is fantastic.