Red Hat Bugzilla – Bug 817958
jenkins benchmark scale-up events hit application limit in an unusual place
Last modified: 2015-05-14 18:53:52 EDT
Description of problem:
When the benchmark job's profiles are analysed, they show that the second and third scale-up events for a scalable jboss app don't run the connector-hooks. This behaviour doesn't repeat on the same image when doing it by hand.
Version-Release number of selected component (if applicable):
Frequently in the Jenkins benchmark job.
Never outside of the jenkins benchmark job.
Steps to Reproduce:
1. Run the benchmark job.
2. Collect the artefacts and analyse with rhc-profiler-merge-report
3. Examine the output for scaled jboss applications.
The second and third scale-up events don't have connector-execute calls.
All scale-up events have connector-execute calls.
Changing severity to low since this only appears to affect benchmark testing and not real application work-flow.
The call gets a 500 error, and we see the following close to the end. Strangely enough though this message is after we talk create the gear and talk to dynect. There's code in the benchmark to increase gears for the test user. This doesn't happen if the app is created and scaled by hand or by a script that runs on its own.
Application limit has reached for 'email@example.com'
Git commit 0ac0135b6 fixes a bad command line argument to increase the max gears per user.
Clearing this bug off as wontfix. Yes its an awkward spot to do it but we no longer return a server error; instead we return a 422 error.