Hide Forgot
This was fixed in 5.6.1. Please move to CLOSED CURRENTRELEASE if verification passes.
From the PR: ======================================================= simplify capture_vm_targets Tangentially related to #8429 and https://bugzilla.redhat.com/show_bug.cgi?id=1332579 https://bugzilla.redhat.com/show_bug.cgi?id=1331803 Before: traversed all vm parents instead of the ones that we said were relevant. [performance] returned vms even if they did not have an ems [bug] After: loop over the parents we said were relevant ensure the vm has an ems before returning it. This is no longer called for cloud vms, and doesn't show up for infra vms. Put this in anyway to ensure it is part of the contract. ========================================================= Keenan, steps to verify? Is this in reports where some VMs were included when they should not of been or is purely just a speed improvement that is going to be be hard to verify?
There are a number of performance related BZs around capacity and utilization capture. The part that puts the targets (storage, host, vm) into the queue to be captured. What you should see. if you have 10k VMs. it will complete the capture_vm_targets job within a minute or so rather than over 10 minutes (and over the timeout threshold). There will no longer be errors in the log. You may want to test all of these BZs in 1 fell swoop. All of them were complaining that this process timed out, and the testing of these BZs is that it did not timeout even though there are many many vms. Let me know if you need more info
tested on 5610: 1 db appliance, 3 worker appliances, 10k vms on vcenter, w/ metrics collector worker on all appliances As expected, completion happens within 30 seconds approx. Script used to extract timings: https://github.com/arcolife/cfme-performance/blob/dev/adhoc-scripts/get_metric_capture_avg.sh