Where do you want me to send logs. I didn't post them here because I read somewhere that this is unwanted.
I'll try to explain what I've done:
One server with VMware ESX 3.0.2 plus Qlogic qla4052c, this is the initiator.
The target consists of a server with Windows Server 2003 plus StarWind 3.5 installed on a PATA drive plus an extra Seagate Cheetah SCSI disk connected to the onboard SCSI controller and a 1Gb NIC dedicated to iSCSI traffic.
When I configure an image file as target I place it on the Cheetah that is formatted with NTFS.
When I configure a SPTI or Diskbridge device I use the Cheetah as a raw device.
Using the qla4052c as initiator I can only connect to an Image File or Diskbridge device.
Using the ESX3.x software initiator I can also connect to the Cheetah as SPTI device.
One question:
You say an image file is faster but is this also true for ESX?
As you probably know a virtual disk on VMFS consists mainly of a large monolitic file (.vmdk) that is allocated the minute you create a vdisk. This file is not fragmented so I can't see why this should be put into another even larger monilitic file being the image file device.
anton (staff) wrote:It's still not clear what exactly did not work. Was it QLogic iSCSI adapter failing to connect to the StarWind-mapped Cheetah drive or what? Where are the logs from this issue? As you've kindly did not provide any
I can only guess it's famous "buffer size" issue (see tape sharing related topic). So in any case please stick with "cached" image file solution as it's absolutely fastest scenario.
Han wrote:I've been experimenting with the combination qla4052c + ESX 3.0.2 -> StarWind 3.5.
Using a Seagate Cheetah in the StarWind server I was:
- Unable to connect to it when configured as SPTI device
- Able to connect when using an image file on that disk
- Able to connect when configured as a diskbridge device
However I was able to connect to the Cheetah when configured as SPTI device if I used the ESX software iSCSI initiator.
Any ideas what could cause this?
Han.
