-
Feature Request
-
Resolution: Unresolved
-
Undefined
-
None
-
2.4
-
None
-
False
-
-
False
Description
Due to the limited space allocated for /tmp, receptor may occasionally consume all available space during JT executions, potentially leading to application or system crashes or unexpected behavior.
Customer Scenario:
In our Automation Controller a job was started for 69169 target hosts by accident. The problem with this was the fact that on the controller node and one execution node the directory "/tmp/receptor/<hostname>/<id>/" included a file "stdout" that showed up with a size of 5 GB. That filled up the partition completely and caused an partial outage of the affected systems and a lot of job failures within the Automation Platform.
Findings:
- We do not have benchmarking results or suggestion in the documentation.
- While testing, we noticed that the temporary JT execution data gets stored in /tmp + /tmp/receptor + the execution node's /tmp and /tmp/receptor(ignore if no external execution node)
- We do have a KCS(https://access.redhat.com/solutions/7036473) that helps configure a different path for the receptor, but that is a reactive approach for the problem described above.
Ask
Looking for proactive solution for this issue.
It can be handled at the receptor level or documentation level.
Doc Level - Clearly mention the receptor requirements
Receptor Level - Provide an option to configure different path at the time of new installation.