Multipath RHEV 3.4

Latest response

Hello Guys

I have a environment with RHEV 3.4, Storage Equallogic.

My iSCSI network setup have 2 NIC with the same segment iSCSI and I configure 2 iSCSI Network in RHEVM called iSCSI01 and iSCSI02 with 1 NIC by Network.

IP iSCSI Storage: 192.168.1.100
NIC P1P1---> iSCSI01 - IP 192.168.1.10
NIC P1P2---> iSCSI02 - IP 192.168.1.111

The communication with Storage is correctly but when a port is shutdown the connection with the storage is lost. When the port is UP the communication with Storage is OK.

I wonder whether to make an additional configuration in RHEV-h for using multipath to avoid problems of loss of communication with the storage

Thank for your help.

Sebastian

Responses

how about to make a bond(team) interface?

I'm not exactly sure, but it sounds like you're having a similar problem we had. There were two different NICs with their own paths (in our case two different switches without any VPC or MLAG support, so we couldn't bond across them) to the SAN, and when one switch went down, the storage domain was marked as offline and all our VMs were paused.

The problem is that the default timeout for paths in the iscsi config files is something like 120 seconds, and at the time on version 3.4 that file couldn't be persisted so we couldn't save a good number to it.

I don't know if this is a good solution or not, but we worked back and forth with Redhat support and as a solution ended up adding this to /etc/rc.local and persisted it:

for x in $(ls /sys/class/iscsi_session); do echo 5 > /sys/class/iscsi_session/$x/recovery_tmo; done

This sets the path timeout to 5 seconds after iscsid has already loaded and created all the paths. You can probably set it as low as 1 second if you're comfortable doing so. Also note, if it fails over for some reason it can go back to the default 120 when it re-connects back after the path repairs, so you'll want to re-run that manually if you know that the path had failed and repaired, such as when doing rolling switch firmware updates, as was our cause of the problem in the first place.

A feature called "iSCSI Bond" for EQL multipathing was added in RHEV 3.4. I have been using it since December.

This feature is mentioned in the RHEV 3.4 Manager Release Notes and in Solution #96293.

Close

Welcome! Check out the Getting Started with Red Hat page for quick tours and guides for common tasks.