ISCSI Support


Recommended Posts

*pokes head inside*

 

I'm waiting on some hardware to arrive - upgrading to a dual-socket board + processors and a buttload of memory. The idea is to replace + consolidate my storage and virtualization boxes into one, saving space/power/whatever. So I have an opportunity to look at changing things up a bunch!

 

Still, a bit sad to see no iSCSI support (or seemingly even not much dev interest), but I understand it's not an overnight affair anyway. I'll check again another year, heh.

 

Even if I could learn to write a plugin, there's no way I would have time to maintain and responsibly test/debug it for consumption. I'd hate to be responsible for downtime or heartache.

But indeed if anyone is interested, LIO (probably the better iSCSI target to adopt, IMO) pretty much involves wrapping around targetcli functions and /sys outputs. If you had that much working, it's a step away from enabling people FCoE and other goodies that LIO offers.

Link to comment
  • 1 month later...

All,

 

I was able to recompile the unraid kernel for Unraid 6.3rc4 to include the necessary headers to get targetcli working on unRAID. I'm still in very early stages of getting things working and learning about iSCSI targets. I'm a bit lost on what benefit you guys are expecting by having unraid as an iSCSI target. I understand the difference between shares and block level access, but the use cases I've seen so far don't seem like they'd gain much from iSCSI. I believe the main arguments are for performance increases, however the performance would still be bottlenecked by the parity write speeds. I believe I'm missing something fundamentally, so anybody can educate me, I'd appreciate it.

 

Link to comment

All,

 

I was able to recompile the unraid kernel for Unraid 6.3rc4 to include the necessary headers to get targetcli working on unRAID. I'm still in very early stages of getting things working and learning about iSCSI targets. I'm a bit lost on what benefit you guys are expecting by having unraid as an iSCSI target. I understand the difference between shares and block level access, but the use cases I've seen so far don't seem like they'd gain much from iSCSI. I believe the main arguments are for performance increases, however the performance would still be bottlenecked by the parity write speeds. I believe I'm missing something fundamentally, so anybody can educate me, I'd appreciate it.

 

Good question.  What's the use case for iSCSI?

Link to comment

One good case would be for home labs, and for IT professionals using LT NAS in conjunction with other system OS's like Microsoft, Citrix, Redhat, and VMware.  ISCSI is a cross platform industry standard for connecting storage and it would be nice to have.

 

iSCSI is a block-level protocol.  How should an iSCSI LUN be mapped to unRAID storage?  The only use case I can see for this is a SAN.

Link to comment

One good case would be for home labs, and for IT professionals using LT NAS in conjunction with other system OS's like Microsoft, Citrix, Redhat, and VMware.  ISCSI is a cross platform industry standard for connecting storage and it would be nice to have.

 

iSCSI is a block-level protocol.  How should an iSCSI LUN be mapped to unRAID storage?  The only use case I can see for this is a SAN.

Probably like Windows Server and Linux do by creating virtual disk image on existing filesystem and using it for iSCSI target. Even if somebody wants to create plugin they need support in kernel.

Link to comment

For those interested in trying out iSCSI and figuring out if it works for you, I have a working install with the proper activated kernel modules. I currently am using targetcli in a docker and it works great. There is no GUI so you have to be comfortable with setting up the iqn, acls, luns, portals, etc. manually.

 

I'm currently using iSCSI to allow guest VMs block level access to drives without having to passthrough drive controllers. One thing to keep in mind is you cannot create block level luns for drives that are already mounted. If you want to use those drives, you'll have to use fileio.

Link to comment

I'm currently using iSCSI to allow guest VMs block level access to drives without having to passthrough drive controllers. One thing to keep in mind is you cannot create block level luns for drives that are already mounted. If you want to use those drives, you'll have to use fileio.

 

You can pass through block devices to guest VMs without using iSCSI.  In fact, its super easy to do.  First, login via SSH or Telnet and type this command:

 

v /dev/disk/by-id

 

Locate the disk you wish to pass through and copy its name.  Now go create your VM and under vDisk location, select "Manual" and in the path field, type /dev/disk/by-id/ and then paste in the name of the disk you wish to pass through.  You can optionally include the partition so if you want, you can have one disk with 3 partitions and pass through different partitions to different VMs if you want.  What does iSCSI gain you over this method?

Link to comment

I'm currently using iSCSI to allow guest VMs block level access to drives without having to passthrough drive controllers. One thing to keep in mind is you cannot create block level luns for drives that are already mounted. If you want to use those drives, you'll have to use fileio.

 

You can pass through block devices to guest VMs without using iSCSI.  In fact, its super easy to do.  First, login via SSH or Telnet and type this command:

 

v /dev/disk/by-id

 

Locate the disk you wish to pass through and copy its name.  Now go create your VM and under vDisk location, select "Manual" and in the path field, type /dev/disk/by-id/ and then paste in the name of the disk you wish to pass through.  You can optionally include the partition so if you want, you can have one disk with 3 partitions and pass through different partitions to different VMs if you want.  What does iSCSI gain you over this method?

 

It gains me nothing for my use case. Your method is likely faster, I just didn't know about it! I'm actually very excited to try that out.

Link to comment

I sense a little resistance, it must be personal when there are so many request from the user base.  I am sure we can work around or find a different platform.  The community is probably a mix of Gamer's and hobbies/home lab IT folks that would like to use similar technology that they either are learning or working with.

Link to comment

I sense a little resistance, it must be personal when there are so many request from the user base.

Very seriously doubt it's personal, just business. The amount of effort to develop and properly support it doesn't look to be worth it compared to the added benefits given the use cases presented right now. The number of people showing interest really is pretty small compared to the people wanting better VM and docker support.

 

Features in unraid tend to be worked on based on perceived need, so if more people want a feature, that's where the limited development and support tend to go. Give it time, and maybe iSCSI's time will come.

 

We lobbied for years to get UPS support, external notifications, dual parity, etc added. Now we have all that plus VM's and dockers.

Link to comment

I understand about being pragmatic,  I am fairly new to Lime Tech / unraid and have 25 years of IT under my belt.  I have used Kimchi, Ovirt, Linux, Xen Server, Citrix, Hyper-V, Vmware/VMUG looking for an all around home lab platform that can be easily used as a game machine on the weekend and home lab during the week.  This what has attracted me to move over to Lime Tech / unraid,  prior to Lime Tech  I was using my VMug subscription to provide the GPU passtru for gaming with VMware and for a lab environment, this seem to be to bloated so I tested Rhev / Ovirt which was nice but didn't like distributive architecture and the number of servers required.  My last venture was with Fedora and using Kimchi which was straight forward KVM/Libvirt with HTML5 interface which used NoVNC and Spice, however it did not offer total platform like Lime Tech.  So from my perspective if Lime Tech had ISCSI, SPice, and VM Cloning bundled in its HTML5 "Single Pain of Glass"  this would be the Ferrari vs a Lexus.  I also give you Kudos on the community interaction,  I use Samsung Smartthings Home Automation and they have very good community support as well which brings on a lot of market share from its competitors which drives more customers of all types. Great job on the software....I wish I came up with the idea.

Link to comment

I sense a little resistance, it must be personal when there are so many request from the user base.  I am sure we can work around or find a different platform.  The community is probably a mix of Gamer's and hobbies/home lab IT folks that would like to use similar technology that they either are learning or working with.

 

From observing the forums I wouldn't agree that there are many users requesting it. And if we only count request accompanied by solid use cases rather than simple "I want it"s, I can almost count them on the fingers of one foot ;)

 

Rather than assuming it is personal, focus on giving some practical use cases. If it is clear what is being asked for and many users want it, I am sure it will be thoroughly considered.

Link to comment

I think jonathanm made his point and I responded.  Use case would be an investment of time to gain greater market share.  since I do not have the cost of development of adding the ISCSI features set to the software,  I am unable to tell you what the financial impact would be or model a ROI / business case.  However if we are looking at bring in more and different user into the user base $$ the method of using what is being post as an "want" in the forum is anecdotal and not really a good measurement in this case. Since the folks that really are looking for the added feature are probably using using some like FreeNas, Nas4Free, openvault etc...to meet their requirement.  And these are the folks that could provide a greater market share / revenue stream, and probably are folks that are the "Home Lab" users that pay for subscriptions like VMUG ~$200 a year. The "Home Lab" market using NAS, Docker, and VM technology is no small market,  considering VMware has invested in VMug to promote there brand. 

Link to comment

Why iSCSI?

-because every comparable NAS solution already has it (FreeNAS, Synology, Qnap, OpenFiler...),

-because it is easy to add,

-because it would help some of us, already using unRAID for everything else, dump other solutions existing for the sole purpose of serving iSCSI targets (think mostly for ESXi home labs),

-a few other edge cases where access to block-level devices is required or preferred over smb or NFS.

 

On the other note I don't get why some people uninterested in the topic try to discourage Tom from looking at it. I don't have any use for e.g. graphics card pass-through to VM, but I don't advocate to stop wasting development time on it.

Link to comment
  • 1 month later...
  • 4 weeks later...
  • 3 weeks later...

I use it for serving cd's & cd images to vm's. Instead of having the xml file specify a specific ISO file or the physical cdrom, I define an iscsi cdrom in the xml the same for all vm's and change CD's at the iscsi server.  I'm still experimenting as to the best way to define the drive in the xml, as it works great in windows VM's, but OSX I'm finding needs a restart (not full power off) to refresh a changed CD. I can install a fresh OSX from an iscsi DVD ISO, but I have to share it as a disk rather than cdrom. The iscsi target doesn't seem to handle HFS+.  I'm using SCST for the iscsi target, and built libvirt 3.1.0 and Qemu 2.8.0 both with the added iscsi support, as well as some kernel modifications that I read someplace were needed. While I'm only interested in the CD images over iscsi at the moment, my VM's could theoretically use iscsi block devices from the server itself, or from some iscsi toaster like a qnap. 

Link to comment
  • 2 weeks later...

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.