Intermittent loss of paths

[Thomas]

Afternoon everyone I’m currently trialling Starwind and I seem to be coming up against a problem maintaining the paths to the iSCSI SAN below is a little description of my hardware

Starwind;

ML110 G5
Server 2003 R2 (with all current updates from last week when installed)
4GB Ram
HP P400
6 x Seagate 15k SAS drives (R10 Storage)
2 x WD 750 Black (OS)
MS iSCSI Service
1 x BCom NIC(management)
1 x Intel ET dual port(iSCSI traffic)

ESXi 4.1.0, 260247;

ML115 G5
8GB Ram
2 x Intel Pro NIC(iSCSI)
1 x BCom NIC(Management and VMs)

LUN;

Disk Bridge
Persistent Reservations
Asynchronous Mode
Write-Through caching
1024


Starwind is setup to listen only on the Intel NICs and in ESXi I have a round robin setup. Any help would be most appreciated.

[Max (staff)]

Hi, you mean you're loosing connection to one of the pathes?
I assume ESXi's discovery has been already set to the IP addresses of 2 iSCSI nic's.
Can you please describe what settings have you specified to the StarWind server while restricting the iSCSI interfaces.
Have you used ACL's for this?

[Thomas]

Hi Max,

Q. you mean you're loosing connection to one of the pathes?
A. Seems both paths take turns at dropping off, sometime its only one at a time and sometimes both go down

Q. I assume ESXi's discovery has been already set to the IP addresses of 2 iSCSI nic's.
A. Yes Dynamic dicovery has been enabled with both IP Addresses

Q. Can you please describe what settings have you specified to the StarWind server while restricting the iSCSI interfaces.
A. I assumed that what the Configuration->Network tab was for (I right clicked the IPs that I don't want iSCSI taffic on and selected disable

Q. Have you used ACL's for this?
A. I removed the swicth and connected the ESXi server and Starwind Box directly via RJ45 to rule out the Switch

[Max (staff)]

Ok, I've got the point!
You need to recreate the device enabling multiple iSCSI connections (clustering) in the device creation wizard.
This will remove one session limitation for the initiator side.

[Thomas]

Sorry Max I should have mentioned that it’s already enabled (both paths are marked active I/O, Round Robin)


[Added]
Here is the log at the time its happening

Code: Select all

9/13 16:15:27.031 884 C[8b], IN_LOGIN: Event - LOGIN_ACCEPT.
9/13 16:15:27.031 884 C[8b], LIN: T5.
9/13 16:25:31.406 844 T[8b,13a]: Management command: abort task (CmdSN 278496, ITT 0xee4c0400) - task not found.
9/13 16:25:49.234 844 T[8b,13b]: recvDataBuf failed.
9/13 16:25:49.234 844 T[8b,13b]: recvScsiData failed (0/8192)!
9/13 16:25:49.234 844 C[8b], LIN: *** 'recv' thread: recv failed 10058.
9/13 16:25:49.234 854 C[8b], LIN: WSASend() returned 10054!
9/13 16:25:49.234 884 Tgt: close 'iqn.2008-08.com.starwindsoftware:iscsi-15lun': 1 session(s) opened, 65535 more allowed.
9/13 16:35:23.140 37c Srv: Accepted iSCSI connection from 10.10.x.x:54190 to 10.10.x.x:3260. (Id = 0x8c)
9/13 16:35:23.140 37c C[8c], FREE: Event - CONNECTED.
9/13 16:35:23.140 37c C[8c], XPT_UP: T3.
9/13 16:35:23.140 b08 C[8c], XPT_UP: Login request: ISID 0x00023d000002, TSIH 0x0000.
9/13 16:35:23.140 b08 C[8c], XPT_UP: Event - LOGIN.
9/13 16:35:23.140 b08 C[8c], IN_LOGIN: T4.

[Constantin (staff)]

Please, send us at support@starwindsoftware.com full log! We`ll take care of it.
Thanks a lot.

[Thomas]

Thanks Constantin, I have done as you requested.

[Constantin (staff)]

Yes, I`m currently watching your logs, and will answer you over email.

[Thomas]

Turns out my 2 month old Intel ET Dual port NIC was broken, a kind thank you to the starwind boys for there help in this matter.

[Constantin (staff)]

Any time