-
Task
-
Resolution: Done
-
Major
-
None
-
None
Description
Bonfire uses "--remove-resources all" option by default. This causes that host-inventory doesn't have enough resources in ephemeral environment, which causes host-inventory-service to restart randomly and reject some requests.
The issue has been fixed on the host-inventory PR check by adding host-inventory to the "COMPONENTS_W_RESOURCES" env variable, but I'm afraid other PR checks might still see the issues, because a lot of different apps PR checks are dependent on Inventory too.
I see 3 options to fix the issue globally:
- Update bonfire to use "--no-remove-resources host-inventory" by default
- Update bonfire to give host-inventory more resources with "--remove-resources host-inventory"
- Stop using "--remove-resources all" by default
I would suggest the last one. I don't think ignoring things from the config files implicitly is a good practice. It can cause a lot of trouble to users which are not familiar with every bonfire option and don't know that this is the default behavior. I think a better way to limit spending would be checking that the services don't use excessive amount of resources in their app-interface config files in combination with some maximum limit for the resources requests.
You can find more details about the issue here: https://issues.redhat.com/browse/RHINENG-4100?focusedId=23718000&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-23718000