Make full-output configurable

[chforster]

I currently have the problem that my Jenkins runs into an "OutOfMemory" Exception when performing a 'cmd.run' on multiple machines which have a big return (e.g. unzip a large file without quiet parameter).
Same occurs when applying states.

It would be nice to make the "full_return" configurable which is currently hardcoded

saltstack-plugin/src/main/java/com/waytta/SaltAPIBuilder.java

Line 382 in cbc6ca5

saltFunc.put("full_return", true);

[mchugh19]

Sure, that makes sense. I'll look into adding that.

[mchugh19]

I've been looking over the output and am not quite sure why you are seeing the OutOfMemory error.

Here's an example of a tar extraction run via the salt-api and cmd.run:

{"return": [{"master": "./
./_modules/
./_modules/saltcheck.py
...
"}]}

Here's the same job run with full_return: "True"

{"return": [{"master": {"jid": "20180829153616106818", "retcode": 0, "ret": "./
./_modules/
./_modules/saltcheck.py
...
"}}]}

I can make a release which disables the full_return option, but it might be in the parsing of the large json return that you are actually seeing the problem. I'll work up a beta release to disable full_return, and we can go from there.

As a quick test, you could try enabling the "Skip validation of salt return" option. This should skip over the processing of any large JSON output.

[chforster]

Currenty I can not reproduce the Issue due to missing system resources.

It happened when applying a state via your Jenkins Plugin NOT async and parallel on 10 machines (wildcard grain), which triggered a shell script (via cmd.run) which then extracted a big tar file (1.2G - approx 50.000 Files).
The log Output was not saved to a specific Workspace File but in Jenkins System Log.

Here my environment Info:

• Jenkins 2.140
• Saltstack Plugin 3.2.0
• salt 2018.3.0 (Oxygen)
• salt-api 2018.3.0 (Oxygen)
• TornadoServer/4.2.1

Configuration in Jenkins:

• SaltAPI Version: 2017.7
• Display format: json

I recently solved the Problem by applying the state directly on the saltmaster and not from Jenkins.

[mchugh19]

If you are able to capture the salt-api output of that operation, could you sent it to me? ([email protected]) That use case could be used to test large json returns.

Also, if you ever run into problems like that in the future, try skipping validation of the return. That is the probable cause of resource utilization.