ESXI trying to connect to ISCSI storage with both storage adapters - session in use

[parisvc]

We have a Vmware 6.0 cluster and using Netapp FAS2240-4 (8.23) for storage.

We've been experiencing disconnects on vm's hard drives.

on the netapp error log I can see this:

ISCSI: New session from initiator iqn.1990-07.com.emulex:dev-svr-03 at IP addr 10.0.0.3

ISCSI: New session from initiator iqn.1990-07.com.emulex:dev-svr-03 at IP addr 10.0.0.19

ISCSI: New session request from initiator iqn.1990-07.com.emulex:dev-svr-03, a session from this initiator already exists.

each esxi host is using hardware iscsi has vmhba1 and vmhba2 configured with an ip address each as above. I'm just wondering why it's trying to connect with both and therefore giving us this error?

[ThompsG]

Hi parisvc

Could you confirm if you are using independent or dependant ISCSI adapters? In both cases I would have expected the iSCSI Name to have been unique between vmhba1 and vmhba2?

From what is posted it appears that both have the same iSCSI Name?

Thanks in advance

[parisvc]

We are using HP ProLiant BL460c with HP FlexFabric 10Gb 2-port 554FLB adapter

[root@dev-svr-03:~] esxcfg-scsidevs -a

vmhba0  hpsa              link-n/a  sas.50014380268127f0                    (0000:03:00.0) Hewlett Packard Enterprise Smart Array P220i

vmhba1  be2iscsi          online    iqn.1990-07.com.emulex:dev-svr-03    (0000:04:00.2) Emulex Corporation Emulex OneConnect OCe11100 10GbE, iSCSI UCNA

vmhba2  be2iscsi          online    iqn.1990-07.com.emulex:dev-svr-03    (0000:04:00.3) Emulex Corporation Emulex OneConnect OCe11100 10GbE, iSCSI UCNA

From reading the below 1 adapter with 2 ports means they use the same iqn....

Two ISCSI HBAs emulex oneconnect with identical IQN and can't multipath

[ThompsG]

Yes - I was thinking about this last night and the software ISCSI device only has one IQN but multiple IP addresses.

Thinking some more about this

[ThompsG]

Hi there,

After a little searching it appears that there might be two things you should check:

1. Is the MTU configured the same on all the ports between the ESXi server and storage? Reference: SDW and Windows 2012 slow enumerating LUNs - NetApp Community and the solution was that not all ports were configured to accept Jumbo Frames
2. Upgrading the Data ONTAP version. Reference: iSCSI connection failure on LACP re-balance - NetApp Community as this indicates that an upgrade to a newer version than you currently have solved it for them.

I would investigate option 1 first before contemplating the other and obviously check your NetApp upgrade matrix to see if this is a possibility.

Kind regards.

[parisvc]

Many thanks for your research.

1. Actually no I posted this post also: esxi and netapp MTU

Do you know if there's anything I can use to test the mtu settings with as I can't vmkping on an iscsi adapter?

2. The latest version of ontap we can use for our Netapps is 8.2.5 so we are going to upgrade from 8.2.3 but it looks like we won't be able to get the version that you have linked.

[ThompsG]

Hi there,

Sorry for the slow response but have been focused elsewhere.

You can actually use vmkping on the iSCSI interface by using the -I option to specify the vmk to use. Here is an article that explains it better than I can: https://kb.vmware.com/s/article/1003728

One thing to note - make sure you use the -d option so that packets cannot be fragmented to fully test.

Let us know how this works out.

[parisvc]

we are using hardware iscsi so I can't vmkping from the iscsi adapter?

[root@dev-svr-03:~] esxcfg-hwiscsi -l vmhba1

DHCP Configuration Disabled

ARP Redirection Disabled

MTU Size = 8342

[root@dev-svr-03:~] vmkping -I vmhba1 -d -s 8342 10.0.5.134

Unknown interface 'vmhba1': Invalid argument

[MikeStoica]

You don't need to put vmhba1 when doing vmkping. Do it again with vmhba1