Special procedure is needed to upgrade HE environment with RHEL6 hosts to RHEL7. Engineering has provided a workaround and QE needs to test it before we advise this to the customers. Note: need to make sure this works with RHEV-H hosts as well. Version-Release number of selected component (if applicable): Upgrade of 3.5 HE hosts based RHEL6 to 3.6. HE 1.2 stands for 3.5 packages. HE 1.3 stands for 3.6 pacakage. Problem: 3.6 Hosted Engine Storage structure has changed from 3.5 to 3.6. (specifically location of vm.conf) If upgrading el7 host, that is already part of HE setup, updating ovirt-hosted-engine-ha package would initiate HE storage migration to a new structure. This will work fine in any scenario, since this host is already part of the setup and knows where the current HE vm.conf is located. If adding new el7 host to HE environment that is based el6 hosts only(regardless of the manager version, thus bz#1311027 is pointless), it will not work. (bz#1336950, for instance). el7 channel contains the latest HE and vdsm packages. So, the new packages would install and point to the new location of vm.conf. And there is no way it can join an existing setup, that still points to the old location of vm.conf. The workaround is to get a fresh el7 host and install HE 1.2 (aka 3.5) packages on it, add the host to the setup and then upgrade to 3.6. The suggested command is : # yum install ovirt-hosted-engine-ha-1.2.10-1.el7ev ovirt-hosted-engine-setup-1.2.6.1-1.el7 As suggests the solution: https://access.redhat.com/solutions/2300331 However, this will bring the newer, 3.6, vdsm packages and this may create some conflicts. That's why we need to have this procedure tested by QE or provide a different procedure. And of course, the docs should be modified. Bz#1328382. P.S. I didn't have the opportunity to review how this works with RHEV-H. Hopefully, if we get last 3.5 RHEV-H based el7, add it to the environment with el6 hosts and then upgrade it to latest RHEV-H el7 with the correct HA package and vm.conf location. Need to make sure this is tested as well.
A procedure is not a bug. Testing request is not a bug either.
Can you please give me a reason why a customer wants to add RHEL7 hosts to the cluster of RHEL6 hosts, why just do not upgrade all hosts to RHEL7? And just to be sure about verification, steps are: 1) Deploy 3.5 HE on RHEL6 host 2) Try to add RHEL7 host with 3.5 packages to HE environment Correct me if I wrong. But again I do not think it is a good idea to mix RHEL6 and RHEL7, also under 3.6 we have InClusterUpgrade scheduler policy, that will make possible to upgrade all 3.5 RHEL6 HE engine environment to 3.6 RHEL7.
*** This bug has been marked as a duplicate of bug 1328382 ***