SSA Fails with PR https://github.com/ManageIQ/manageiq/issues/14356

[jerryk55]

When we attempt to run SSA after PR #14356
is merged, there is no exception but SSA simply does not work. It starts to run, then says its finished, but never does anything. The same thing happens with both VMware and SCVMM/Hyper-V. If you look at the Tasks Page in the UI - the SSA tasks have a state of “finished”, a message of “scanning for metadata from VM”, and an empty Owner Message. I haven’t found the reason in the log yet, but I was able to reproduce this several times when I was attempting to test VDDK 6.5 unsuccessfully. I ran into the same thing on the microsoft platform. No data is found for the VM because no actual analysis has been run. If the above PR is (manually) backed out, SSA is back to normal.

[Fryguy]

Note that if we have to revert #14356 then we will also have to revert #14380 . Reverting the migration could get complicated, so if we can solve this with a simple code fix, that would be preferable.

[jerryk55]

Agreed. But the if code fix requires using jobs.agent_class, we are SOL without reverting the migration anyhow.

[Fryguy]

According to @yrudman, the only value that ever appeared in agent_class was MiqServer, so ideally, the code could be adjusted accordingly. I'm still surprised you don't have a backtrace anywhere.

[jerryk55]

Ok I did just find a backtrace but I'm not 100% sure its related. I have a couple of other backtraces in my log that I know are not related - one is re: OpenStack and Ironic not available, and one is for the Microsoft Refresh code because of an WMI open connections issue that @djberg96 has open. They are not for scans. I just came across two backtraces - one isn't listed as an ERROR - its a Deprecation error in VmOrTemplate::ems_host_list. The second is in call_scan - I think this is possibly the issue -
[----] E, [2017-03-21T10:28:47.736012 #30918:531998] ERROR -- : MIQ(VmScan#call_scan) undefined local variable or method host' for #<VmScan:0x0000000c901bc0> [----] E, [2017-03-21T10:28:47.784121 #30918:531998] ERROR -- : MIQ(VmScan#call_scan) /opt/rubies/ruby-2.2.4/lib/ruby/gems/2.2.0/gems/activemodel-5.0.2/lib/active_model/attribute_methods.rb:433:in method_missing'
/mnt/hgfs/jerryk/redhat/manageiq/app/models/vm_scan.rb:168:in call_scan' /mnt/hgfs/jerryk/redhat/manageiq/app/models/vm_scan.rb:519:in snapshot_complete'
/mnt/hgfs/jerryk/redhat/manageiq/app/models/job/state_machine.rb:34:in signal' /mnt/hgfs/jerryk/redhat/manageiq/app/models/vm_scan.rb:126:in call_snapshot_create'
/mnt/hgfs/jerryk/redhat/manageiq/app/models/job/state_machine.rb:34:in signal' /mnt/hgfs/jerryk/redhat/manageiq/app/models/vm_scan.rb:85:in check_policy_complete

(there's more)

[yrudman]

I did tested original big PR (removed all references to agent_class, agent_state, agent_id but table structure was the same) in 2 environmentsL VmWare and RHEL

SSA was finished successfully with updated numbers on users, packages, etc ...

[yrudman]

OK, i found issue

[yrudman]

PR: #14444

[jerryk55]

#14444 (review) fixes this issue. When it is merged we can close the issue.