Cloudmin Master cannot initiate using itself as an iSCSI server

If you try and choose the cloudmin master in the Add iSCSI server: dialog it goes to white screen with the Iscsi Servers header and never completes, no errors are given and the server cannot be used.



Can I login to your system again to take a look at this?

Sadly our infrastructure is all down as a forest fire burned the fiber :( should be up tomorrow, and I will let you know as soon as its up so you can have a peek.

ok we are back up, comms.virtual is a good target to try, when it fails it leaves all the iSCSI infrastructure behind so on subsequent efforts it will fail saying that harddrives already exist.

Oops that last comment was for another thread, But jamie you are good to go to login and try setting cloudmin master as an iSCSI server..... Franco

I found a bug that could cause t his, and fixed it on on your system. Let me know if you are still seeing the issue.

Sweet, another bug solved :) works now!

I found a further problem related to this. (more of the same bug or perhaps a differnt one)

After selecting the cloudmin master an an iSCSI server and choosing lg_master as the LVG it will use for iSCSI, I then later tried to move an iSCSI disk from SMVMCLOUD to SMVMHOST and it failed:

Moving disk for /dev/vda to storage iSCSI server smvmhost .......
.. move failed : Failed to create new disk : LVM logical volume creation failed : Failed to read from fastrpc.cgi :

during the process instead of using the chosen/defined LG (lg_master) it created a new one instead. I think this is a bug as it should build the logical volume in the selected LVG instead of creating a new LVG.

Retrying the move to a different server (From SMVMCLOUD to SMVMKVM) also failed:

Moving disk for /dev/vda to storage iSCSI server smvmkvm.hardware .......
.. move failed : Failed to create new disk : LVM logical volume creation failed : Failed to read from fastrpc.cgi :

But had a different effect on the server, it actually made the LV in the correct LVG as was defined by its iSCSI server param so it used lg_iscsi correctly.

Are your Cloudmin master system and the VM hosts all directly connected on the same LAN, or is there any firewall or router in between them? I ask because it looks like a lot of these failures are due to connections being lost between systems, and the solutions I've applied are really just work-arounds.

yes all directly connected. They are all on the same segment of a private gigabit network dedicated to the 4 KVM servers and the other physical system that support them.

Actually it just dawned on me, that this could be a problem with iSCSI and IP addresses. Three of the 4 KVM servers have have mutiple NICs and Bridges, what if this is related to the iSCSI subsystem engaging the wrong IP? I have seen it grab instead of the 192.169.80.x address when dynamically building iSCSI conections, I have also seen it create the connections with the 216.108.x.x address as well....

I now have wireshark set up on this segment, perhaps we can use it to debug the problem, I can run it to capture traffic while I attempt a move, what kind of traffic would it be?

If you are seeing multiple different IPs used, this may be a routing issue.

Do the hostnames of all your physical systems resolve to addresses on your internal LAN?

yes they do.