Unsolved
This post is more than 5 years old
1 Rookie
•
9 Posts
0
10813
October 8th, 2013 12:00
EQL HitKit not populating MPIO tab
We have had an odd issue affecting 2 of our file clusters (2 node cluster). We noticed within the group manager that the volumes associates with these clusters were only showing one connection per node. What we generally see with MPIO working is 3 connections per volume, per NIC (we have 2 1GB NICs per server for multipath). So when working properly, I see 6 connections to a volume from one server.
What I also noticed is that when I open the iSCSI Initiator in Windows on these affected servers, and I open the Dell Equallogic MPIO tab, the MPIO record list never populates. I just see a message at the top which says "MPIO records are loading, please wait".
I have a case open with Dell but all they have recommended so far is upgrading to the latest version of the HitKit (4.6) which I have done, to no avail. Has anyone else seen this before? Thanks in advance!
CosmoJoe
1 Rookie
•
9 Posts
0
October 9th, 2013 07:00
Thanks for the reply. Actually, the issue is we see this behavior on the active node as well as the passive node :( We had a switch failure recently which took down one of the paths to our EQL SAN. One of our clusters went down hard, losing access to all its disks. As a post mortem review, we noticed this behavior with the cluster and noticed in the Group Manager, the connections to the volumes were showing the same way as described above for both active and passive node.
Joe S586
9 Technologist
•
729 Posts
0
October 9th, 2013 07:00
This could happen when you reboot a cluster node, and it rejoins the cluster as a passive node, it will only create a single path to each cluster resource volume. This is expected. As a passive cluster node, the server is blocked from doing anything with the disk devices, so the EQL MPIO module doesn't have enough information to create all appropriate connections.
However, if you move a resource group over to this node, the node will then be able to recognize the disk(s) in that group as being appropriate for MPIO, and will create the additional paths appropriately.
Note that once a server has been the active node for a resource group, it will keep all the MPIO connections even if it later becomes a passive node, as long as it does not reboot. The single-path behavior only occurs when a node's initial connections to its volumes come when it is a passive node in a cluster. Once it has the full set of paths, it does not abandon them simple because it fails over resource groups.
When support gets back to you, let us know what they found!
-joe
Joe S586
9 Technologist
•
729 Posts
0
October 9th, 2013 08:00
Sounds like a device isn't passing traffic/protocol... test the paths:
Ping and/or traceroute from each iSCSI port from the host to each eth port of the array as follows:
From the Host use;
ping -S host_ip array_member1_eth0 (repeat for all combinations)
tracert -S host_ip array_member1_eth0 (repeat for all combinations of members and eth interfaces)
From the array (login as grpadmin, and telnet to each member in your group (use the eth0 interface IP address not your group IP, if using a management only setup use the management ip address of the member),
To ping from the array:
GroupName>ping "-I "
(that is a –I as in Capital letter “eye”, AND ensure you use the quotes after ping and the end of the command!).
(the dest IP is each of the Host iSCSI interfaces, Test ALL member eth interfaces, test all IP combinations (Group IP; Member1 eth0, eth1, eth2, etc.; Member2 eth0, eth1, eth2 etc.)
Traceroute from the array:
GrpName>support traceroute -s [ETH port source IP] [destinationIP]
(make sure you test each Array ETH/Host INT combination)