Infrastructure Software

Hi all,

I've been tasked to attach an iSCSI accessed LUN on our EMC Clariion CX3-20i to a new Solaris 10u4 x86 system. I'm running into a perplexing issue that seem to be directly related to the initialization of a zpool upon boot.

When the system boots, there is about a 10 minute period of warnings regarding the initiator being unable to connect.

*iscsi: [ID 286457 kern.notice] NOTICE: iscsi connection(13) unable to connect to target iqn.1992-04.com.emc:cx.apm00065006236.b0 (errno:145)*

This occurs after all nics are available and plumbed. It's repeated in logs for about 12 minutes the boot continues.

During normal operation, I also see the following error repeated several times - though I can find no corollary error on our Clariion for either message.

*iscsi: [ID 372941 kern.warning] WARNING: iscsi connection(13) login failed - Target has insufficient session, connection or other resources. (0x03/0x02)*

The client is a Sunfire X2100 using the Broadcom gig nic. It's switchport is set with spanning-tree portfast.

The iSCSI config is pretty basic. Two static-config entries for the EMC in TPGT 1 and 2. The initiator is configured to use two sessions per connection per some MPxIO notes found in various Sun docs.

iscsiadm list target
Target: iqn.1992-04.com.emc:cx.apm00065006236.b0
Alias: -
TPGT: 2
ISID: 4000002a0001
Connections: 0
Target: iqn.1992-04.com.emc:cx.apm00065006236.b0
Alias: 6236.b0
TPGT: 2
ISID: 4000002a0000
Connections: 1
Target: iqn.1992-04.com.emc:cx.apm00065006236.a0
Alias: 6236.a0
TPGT: 1
ISID: 4000002a0001
Connections: 1
Target: iqn.1992-04.com.emc:cx.apm00065006236.a0
Alias: -
TPGT: 1
ISID: 4000002a0000
Connections: 0

With just iSCSI with multi-pathing setup boot times are normal. When a volume is setup with zpool using all defaults:

*zpool create emc disk*

the boot behavior appears.

It really feels like this is a race condition. I've just tried setting the mount point to legacy with no change. This is why I'm assuming something borks during startup config of all things ZFS.

Does this sound like a new bug with the iSCSI implementation? Or can someone point out a config problem I have overlooked?

Thanks