Drobo

Multiple iSCSI Clients

Hello all,

Ok, here’s my plan. I want to have multiple smart volumes on my DroboPro. These will include a 16TB NTFS partition (Accessed from Vista), another 16TB NTFS(Accessed from Win2008Server, maybe even 2 Windows2008Servers running cluster mode), a 16TB HFS+ partition (Accessed from OSX Server), a 2TB ext3 (Accessed from Linux), and an 8TB ext3 or ext4(Accessed from Linux).

I have read the ESXi Best practices document(which seems to be the only document giving advice on multiple clients), but it leaves me with all kinds of questions, like these:

  1. The Best Practices document describes doing a one-time setup of the DroboPro via USB and then from what I can tell, the DroboPro from that point on is not connected to any dashboard. That means there are no notifications or anything in the case of problems? Is there a way to keep a dashboard on this thing? You can read my other posts on this forum to get an idea of my opinion of the lack of information coming out of the DroboPro.

  2. It is stated that port 3260 is for a single client, but 3261 is for multiple clients. But how does this actually work? If I want multiple clients, do I hook the first one (the one with the dashboard) up to port 3260 and the rest to 3261? Or just put all of them on 3261? Will the dashboard work on port 3261? Can I even use a dashboard when I have multiple clients?

  3. Does the DroboPro Mac iSCSI client have any way to configure it to use port 3261?

I’m going to stumble blindly ahead if I don’t get any help, and I’ll let you know how it goes. But if anyone out there has blazed this trail before me already, I’d be interested in hearing your tips…[hr]
Well, here’s what I have so far:

First, I set up a 16TB NTFS, a 2TB NTFS, and an 8TB NTFS on the DroboPro using dashboard on Vista. Then I went into Disk administrator and disabled the 2TB and 8TB volumes. So Vista now just shows me the 16TB volume.

Then, I went into the iSCSI Control Panel and removed the entry for the DroboPro that was operating on port 3260, and replaced it with one on port 3261. After a reboot, I once again had access to my 16TB volume, and the 2TB and 8TB are still disabled.

Then, I started up my OpenSuSE 11.1 machine, and used YaST to set up an iSCSI client connection to the DroboPro, being careful to specify port 3261. All three volumes became visible. So I did mke2fs -j on the 2TB volume. It is still running, and my network graph shows a pretty constant stream of about 15MB/sec as it creates the ext3 filesystem.

While this filesystem is creating, I have gone back to the Vista machine and created and removed a couple folders on the 16TB NTFS volume, which works fine.

So, I have both Linux and Windows Vista accessing separate volumes on the DroboPro, with no crashing. So far so good. More demanding performance testing will be coming soon.

… oh, and as soon as I switched Vista iSCSI to port 3261, the Drobo Dashboard fails to connect to the DroboPro.

Well,

So far the DroboPro is standing up to the abuse I’m laying on it. I just wrote a 30Gig file from Linux, and it completed without problems, posting an average write speed of 55.8MB/s. Reading back the file produced an average read speed of 58.7MB/s. I would say that this is a reasonably good result from the DroboPro, as here are some other numbers for comparison, all done from the exact same Linux client (first number is write speed, and second is read speed):

55.8MB/s 58.7MB/s DroboPro firmware 1.1.3 iSCSI
40.2MB/s 64.3MB/s Netgear ReadyNAS Pro iSCSI
40.1MB/s 92.3MB/s QNAP TS-639Pro iSCSI
80.1MB/s 84.0MB/s Gentoo64 - custom RAID0 iSCSI system
36.7MB/s 33.0MB/s Gentoo64 - custom RAID1 iSCSI system

Most importantly, this is the first time I’ve managed to complete an iSCSI file copy of ANY size at all, from Linux to the DroboPro. So this is looking quite promising!

Next Up - repeating the test simultaneously from a Windows and Linux client at the same time. Let’s kick it up a notch!

Tks Corndog for the detailed recap of ur DroboPro MPIO testing & knowledge sharing! Both are much needed cuz so far we’ve been unable to get our Pro supporting MPIO and we are just playing w/ nothing but Windows Servers 2003/208 & XP & Win7. I should study ur config carefully & apply that in our environment.

Another update:

Now I am copying 1Terabyte of media files onto an 8TB ext3 volume on the DroboPro using OpenSuSE Linux iSCSI, and at the same time copying 100GB of videos onto a 16TB NTFS volume using Vista. If I pull the front plate off my DroboPro and put my ear up to the front of the drives I hear a cacaphony of busyness - they are working VERY hard. This has been going on for about an hour now, and no errors so far. The total of both write operations is still around 55MB/sec[hr]
I must say, if this continues, I’m going to have to change my signature to something a little more positive! Firmware 1.1.3 so far seems to be a real keeper!

Sadly,

After copying nearly 100G the DroboPro disappeared from both client machines. I am now trying to shut both down to see if the DroboPro will recover without intervention like a hard shutdown.

This is a great thread! Regarding running a dashboard over iSCSI. Once the iSCSI is set up, there should be a
local disk device defined ( /dev/sdz ). The linux dashboard should detect the drobo, and provide all the normal
functionality.

The first USB connection is just to setup initial the network config for iSCSI, once that is setup, the linux dashboard has no need for it.

philobyte,

Right. Ok, I was trying to use the Windows dashboard from Vista. I will try your linux dashboard and see if that works. First, I’m seeing what state my DroboPro is in after that failure during the big copy. I’m also aware that it failed while accessing an 8TB ext3 volume, so I’ll try again with my 2TB volume, to see if that has anything to do with it.

I’ll have more updates as my testing continues.

Update:

Well, the NTFS volume was still intact, as was the 2TB ext3 volume. The 8TB volume, however, was broken so badly that Linux won’t even recognize it as an ext3 filesystem any more. Only option is to reformat. So for now I’m going to concentrate my testing on the 2TB volume.

To be fair, this is the first time I’ve EVER made a >2TB ext3 volume on anything, so I’m not even sure if the Linux support (or Open iSCSI software support for that matter) of >2TB ext3 is even stable, having never tested it on other systems. So I can’t really criticize that one.

Testing continues on 2TB. Once it surpasses 200 or 300G I’ll start up a simultaneous copy on the NTFS again, to make the DroboPro glow red.

When the Linux file copy to the 2TB ext3 volume hit 200Gigabytes (total copy is 1TB in size) I started copying 260GB of files onto the 16TB volume from Vista.

Both copies are now proceeding simultaneously, and the Linux file copy has now reached 420GB and the Windows copy has reached 200GB. The total throughput to the DroboPro over this whole simultaneous copy is around 42MB/sec.

I will repeat my copy tests above, that gave 55 and 58MB/sec write and read, after these large copies are done, to see how the system responds when it is starting to fill up. I have a mix of all kinds of drives in the drobopro, with dual disk redundancy enabled, and a total useable space of 3TB. So after these copies are done, it will be nearly half full.

Test completed successfully. The DroboPro worked flawlessly with Linux writing 1TB of data to a 2TB ext3 volume, while Windows Vista simultaneously wrote 270GB of data to a 16TB NTFS volume.

I did some examining of my failed 8TB ext3 volume, and it looks like OpenSuSE did something weird. It created a 2TB partition in the partition table but then formatted it as 8TB. So it really looks like the finger points to Linux for this failure. I’m going to switch to Gentoo Linux, which I can keep up to date with the latest partition editors, kernel, and file system support tools, to make sure I “really” create a true 8TB drive and do it properly. Then I’ll repeat the test again.