TOPIC:

It seems that the DRBD role is transitioning to Primary after TGT has started. This ultimately breaks access to the storage on the new master.

I will try to get an updated version prepared tonight for you to try. This would prevent TGT from loading until the new Master has successfully transitioned the DRBD role to Primary. This should address the main issue you are reporting.

We made a change to deal with your role change scenario. Please test with the following version.
halizard.org/release/iscsi-ha/iscsi-ha-1.5.7.tgz

- unpack the tarball
- move into the scripts directory
- run the installer on top of your existing setup - follow the prompts.

Please post your results. We were able to reproduce the issue and then validated the fix which worked as expected.

Its works! You've made my day!

Thank You very much

I have followed this thread as I have similar problems with the instructions given in "iscsi-ha 2-node cluster howto_1.5_final, Appendix B".

As soon as I have transfered the Windows Server 2k8SP1 VM to the Slave and put the master into maintenance, the storage gets broken, i.e. the 10.10.10.3 floating IP is gone. The CPU usage of the VM raises to 100% and no access is possible.

Only force shutdown the VM will help. I finialize everything including rejoining the master and changing the roles, bring the slave to maintenance and reboot, everything returns to normal good conditions.

I have done this now twice with the same result. Thats why I have upgraded to

halizard.org/release/iscsi-ha/iscsi-ha-1.5.7.tgz

but it has nothing changed for me. The same behaviour and same problem was applicable. (I have answered the install script to keep my existing configuration.)

I am next week out with no access to internet but could further test and report the week after.

Best regards
Andreas

In your scenario, are you promoting slave to master after putting the master in maintenance mode?

If so, please post /var/log/messages from the slave when you are able to test

I do promote slave to pool master via XenCenter as I am asked to do so when I put the master into maintainance mode.

The log file is attached from the last attempt with the updated release 1.5.7.

Logfile sent via e-mail.