Starwind Target becomes unresponsive

[thewafflecaust]

Hi,

I have been trying to set up a small home lab using the free version of the Starwind iSCSI target and have run into a strange issue. The target will become unresponsive whenever more than one ESXi host is connected and some load is placed on it from both hosts.

It will stay responsive and perform fine with a single host connected for an indefinite period, but with both hosts connected and both hosts trying to actively access the target it will freeze to the point where the starwind service wont stop/respond, process needs to be manually killed. The issue only started to occur when I added the second host.

The starwind logs are full of aborted tasks events..

Have I missed something really basic?

Starwind is installed Server 2008 R2 SP1, it presents 1 target (allow clustering box is ticked) with 3 deduplication devices.

The ESXi hosts are ESXi 5 using the software iSCSI initiator.

[Phan7]

Are you using a single target for both hosts to access at the same time? Obviously this would have the cluster option set on it, if so.

[Anatoly (staff)]

Hi,

OK, we have some question:
-what StarWind build and version are you using?
-what target typer are you using (Image File, CDP, DD, etc.)?
-have you created multi-LUN target (one target with different devices)
-how exactly targets are used (what for? what is storing on them)?
-can we see your network topology with all iscsi related connections, switches and bandwidths?

Thank you

[thewafflecaust]

Hi,

OK, we have some question:
-what StarWind build and version are you using?
-what target typer are you using (Image File, CDP, DD, etc.)?
-have you created multi-LUN target (one target with different devices)
-how exactly targets are used (what for? what is storing on them)?
-can we see your network topology with all iscsi related connections, switches and bandwidths?

Thank you

what StarWind build and version are you using? 5.8.1964 Build 20120307
what target typer are you using? DD i think? That means dedupliation device right?
have you created multi-LUN target? Yes, a single target has 3 deduplication devices attached to it.
how exactly targets are used? VMFS Datastores for virtual machines
can we see your network topology with all iscsi related connections, switches and bandwidths? Starwind Box, 2xESXi 5 boxes are all connected to a Gigabit switch with a single link each

[anton (staff)]

Can you send both configuration file and log (zipped please) to support@starwindsoftware.com so we could take a look? Thanks!

[thewafflecaust]

Done that.. guess I just play the waiting game with support now

[anton (staff)]

I'll kick some butts to speed up the whole process just a bit

[thewafflecaust]

Thankyou

[Anatoly (staff)]

I see that your logs are from 2nd of May to the 8th. Would you be so kind to clarify when exactly the issue occurred?

[thewafflecaust]

It happens multiple times in that timeframe.

I just grabbed the entire log folder sorry.

Just looking through the logs:
5/8 21:38:17
5/8 9:10:52
5/7 22:41:42
5/7 21:15:40
5/2 20:27:05
5/2 20:36:10
5/2 19:04:58

[Anatoly (staff)]

OK, we`ve got one question - are your ESX are clustered?

[thewafflecaust]

Yes

[thewafflecaust]

Did some more testing to rule out issues caused by hardware/network layers.

Rebuilt the starwind box from scratch..

Repeating the test with XenServer6 as the host, same result, only difference was that there were no errors in the starwind log file this time. VMs still froze when both hosts started to actively access the the target.

I tried again using this time using 3 targets and a single dedup disk per target. Same result again.

I'm running the microsoft target right now, seems to be running okay with clustered esxi5 hosts.

[Anatoly (staff)]

OK, can you try to run non-Dedup target instead (Baisc Image file for example)?

[thewafflecaust]

Looking good so far using non dedupe devices.... They've only been clustered for about 20 minutes though..

Any thoughts on what might be causing it>?