iSCSI issues with boot and file system creation
807557Feb 20 2007 — edited Jul 27 2007We are having problems getting our Solaris 10 based iSCSI initiator working reliably with our HP iSCSI Feature Pack target. After a fair bit of testing using a test server I'm still not sure where the problem lies but I'll paint the picture first and then hopefully someone can tell whats wrong
Environment
Network 1
- iSCSI Target Server - HP Proliant DL380 G4 Storage Server running Windows 2003 Storage Server with the iSCSI Feature Pack v1.53 (build 268)
- iSCSI Initiator Server1(test) - Sun Sunfire 280R running Solaris 10 11/06 with iSCSI Device Driver patch 119090-22 (latest available AFAIK)
- iSCSI Initiator Server 2 (prod) - Sun Sunfire T2000 running Solaris 10 6/06 with iSCSI Device Driver patch 119090-19
- iSCSI Initiator Server 3 (prod) - HP DL380 G4 running Solaris 10 1/06 with iSCSI Device Driver patch 119091-14
Network 2
- iSCSI Target Server - HP Proliant DL380 G4 Storage Server running Windows 2003 Storage Server with the iSCSI Feature Pack v2.01 (build 456)
i- SCSI Initiator Server 1 (prod) - Sun Sunfire T2000 running Solaris 10 6/06 with iSCSI Device Driver patch 119090-19
Configuration
All Logical Unit's (LU) are presented to each host from the same Storage Pool on the Target Server.
Issues
The issues are all on Network 1 when connecting Servers 1 & 2 to the iSCSI Target Server. The issue is not when trying to connect to the LU's presented to each host, this works fine. When the Initiator Target Discovery (automatic or static) is configured and enabled on either host, regardless of whether i have LU's presented to these hosts or not, when I reboot Servers 1 & 2 on Network 1, each server will hang halfway during boot. The way I fix this is I boot into single user mode from the Solaris cdrom and rename or delete the iSCSI configuration files in the /etc/iscsi directory. The server will then boot but as soon as the Initiator is configured again the boot problem reoccurs
Now this wouldn't be such a problem if all that was affected was bootup but if when LU's are connected to the host I run the format command and label them which is fine. I then try to create a File System (FS) and i get errors and the FS is not created. We use Vrts Storage Foundation 4.1 to create the FS but I have also tried the newfs command within Solaris and this also fails. The error I get when using Vrts SF to create a FS is:-
vxfs mkfs ERROR v-3-25497 Write failure
vxfs mkfs ERROR v-3-21084 wvxfs: bad file number
What I have done
- I have checked all the networking and DNS configuration for each host to make sure this is not the issue.
- We have almost an identical setup on Network 2 in terms of hardware and software on the initiator and the iSCSI connections are working fine. The only difference is the iSCSI Feature Pack on the Target on Network 2 is a later version than on the Target Server on Network 1
- Server 3 on Network 1 is using storage presented to it from the Target Server without any problems. We also have a couple of windows hosts using storage presented from the Target Server with no problems.
- I have tried configuring automatic target discovery and static target discovery on each host but the same symptoms are displayed regardless.
I- have tried presenting LU's of differing sizes but still no luck
My Conclusion
I think there may be an incompatibility between the later versions of the Sun iSCSI initiator and the older version of the HP iSCSI feature pack we are running on Network 2. I'm happy to be corrected but I have come to this conclusion due to the fact Network 2 is working fine despite almost identical equipment and software to Network 1 with the only difference being iSCSI Feature pack version. Also the fact Server 3 in Network 1 is working fine suggests there is nothing wrong with the hardware or iSCSI configuraion in general. But like I said I am open to opinion on this problem so if anyone has other ideas then please let me know.
Any help in resolving this issue would be greatly appreciated.