So here’s the deelio…
Both Lio-target 3.2, 3.4, and 4.0rc always end up causing a CPU stall on CentOS within 6 hours of a single ESXi host using it. I could get it to bug out quicker while performing heavy usage.
Right now I’m testing it out on Ubuntu, it’s already looking more stable.
If anyone knows why it’s soo buggy on CentOS, PLEASE leave a comment, I’d really like to get it running.
Also had a bunch of problems compiling lio-utils both on CentOS and Ubuntu. If you use git, and install 3.0 it will install on CentOS, then u can install the newer version after. However installing the latest stable doesn’t work, comes up with a bunch of errors.
I’ll be posting another blog entry in regards to running Lio_Target on Ubuntu 10.10 within the week!
Update:
PLEASE VISIT http://www.stephenwagner.com/?p=300 for an updated tutorial on how to get Lio-Target running stable on CentOS 6!
rather than sticking with lio-target why not give scst a go ?
there is an interesting thread between the scst developer & kernel maintainer on lio-target vs scst topic, which leads me to try out scst:
http://linux.derkeiler.com/Mailing-Lists/Kernel/2008-12/msg05957.html
http://scst.sourceforge.net/comparison.html
i am still testing but so far it looks better than any oss iscsi target ive tried. at bare minimum, it comes out-of-box with scientific linux 5 thus binary compatible with centos 5, which is a huge plus !
Hi osst,
I have to apologize, although I do dable in iSCSI quite a bit, I don’t profess to be an expert in the field, nor the technology…
I took a look at those links that you posted up. I’ve seen the functionality chart (the second link) before, and from what I understand (I could be wrong), I don’t think it’s really up to date.
When I first set out to find an open-source iSCSI target that could run on Linux, I looked at a few things:
1. Performance
2. Stability
3. SCSI Reservations
At the beginning I was going to be trying both, SCST and Lio. I can’t remember what happened (it was 4-5 months ago), but something set me off from SCST. I ended up investing quite a bit of my time in getting a SAN setup using Lio-Target.
I do alot of virtualization both for internal to my business, and also for clients, and I know that Lio supports the proper SCSI reservations standard for multiple ESXi hosts. There’s actually quite a bit of SAN/NAS solutions out there that already support VMware vSphere, and actually have the VMware stamp of supportable storage hardware.
In my opinion, Lio looked the most promising. And I just recently read that it’s going to be included in to the 2.6.38 kernel which is great news 🙂 I’m not saying it’s better, I’m just saying it’s what worked best for me, and what impressed me the most.
PS, sorry it took so long to approve your comment!
Happy new year!
Stephen
Hi, Stephen,
I use lio-core-3.4 and lio-utils-3.0 in the Ubuntu10.10,
but I still encounter the following error message after run “tcm_dump –s”
======================
Unable to access tcm_root: /sys/kernel/config/target/core
======================
Before running “tcm_dump –s”, I execute the following commands:
======================
cd /lib/modules/2.6.34.kernel/fs/configfs
insmod configfs.ko
cd /lib/modules/2.6.34.kernel/drivers/target
insmod target_core_mod.ko
cd lio-target
insmod iscsi_target_mod.ko
======================
I think that it has some bugs in lio-utils-3.0.
Do you have any suggestions for the problem “Unable to access tcm_root: /sys/kernel/config/target/core”?
Thank you.
BR,
Wenko
Hi Wenko,
Instead of loading all those kernel modules with insmod, forget those and just issue an “/etc/init.d/target start”.
Theoretically, this should start lio and load all applicable modules. Any modules it doesn’t load, should be loaded automatically when issuing configuration commands (such as lio_node and tcm_node).
As for the error message about “Unable to access tcm_root”. This is something numerous other people have had issues with. Back when I first started tinkering with lio, I resolved this by building lio and using older lio-utils.
I recently built a new storage box, and tried numerous combo’s (using older lio-utils, etc…) and could not get away from this error, as of right now I havn’t found a way to get it running on CentOS.
If I have time today, I’ll shoot off a note to RisingTide. I don’t think this is a bug… I think I’m actually just doing something wrong… I’m not sure, but I read some notes mentioning that lio now uses a different way of configuration.
I’ll post results when I get some 🙂
Thanks,
Stephen
Hi, Stephen,
I already setup the LIO in the following environment.
lio-core-3.4 ＋ the newest lio-utils＋ Ubuntu10.10
But I encounter another questions.
When I try to use Microsoft Initiator (version: 2.08) to login the LIO target,
I get the “Service Unavailable” response from LIO target and can not login the LIO target.
Did you ever encounter the question? What are your suggestions for this question?
Thank you.
BR,
Wenko
Hi Wenko,
Few questions just to make sure we are on the same page:
1) Is the target started? (This can be done by running /etc/init.d/target start
2) Did you configure the target? (using tcm_node, lio_node)?
3) Did you setup the initiator ACLs and allow access?
4) Did you add the target portal to view the targets? Or are you directly trying to access it without?
Stephen
Dear Stephen,
After executing the 4 steps that you recommend, the LIO can work now in Ubuntu 10.10.
Thank you.
Best regards,
Wenko
Great to hear! Glad I could help!
Hi Stephen,
I have a problem with tcm_node. I install lio-core and lio-utils, rebuid kernel. Everything was OK. But when I want turn on target /etc/init.d/target start
I have some thing like that:
File “/usr/sbin/tcm_node”, line 26
with open(filename) as f:
^
SyntaxError: invalid syntax
Could You have any idea what is wrong?
I’m having the error when start the target
File “/usr/sbin/tcm_node”, line 26
and when i create a new fileio
[root@localhost home]# tcm_node –fileio fileio_0/LUN0 /home/LUN0 512M
File “/usr/sbin/tcm_node”, line 26
with open(filename) as f:
^
SyntaxError: invalid syntax
[root@localhost home]#