Create a Python program to auto-generate the four kafka.json files in the 0.8.2 format #7
Conversation
… the 0.8.2 format. Auto-generating them should reduce the potential for human error in keeping them in sync.
|
Hey Jeremy, I thought it would be more appropriate to assign this to ekilby, but his name doesn't seem to be available in the dropdown. Any idea why? Also, I've created a fake Pull Request 8 for the purpose of showing the diffs between these files and 0.8.2 Finally, I haven't perfected all the jmxtrans names. There are some that I still have questions on (especially ones involving the '-' or '*' characters) |
|
I like this, and it seems like it could be very close to being generic. Maybe make a "%s-beans.json" with the bean definitions (path, attributes, prefix) and then a generate.py that takes the %s as an arg. So passing kafka-0.8.2 would generate the 4 versions of the kafka-0.8.2.json This file happens to have one attribute per bean, but that isn't generally the case. The * are pretty strange with jmxtrans. There is a way (look for typeNames in their samples) to break the wildcards out into separate streams but then we have no way to make our metadata do the wildcarding on them on the receiving end. This is one of the things that the agent JMX plugin should presumably help with. The way it works now you end up with one stream with all the matched metrics mashed together. This is good when you have one and the name changes, but very bad when you have multiple (see Tomcat Connectors). The hyphens may just need the quote escaping (like the "before" in your other PR) to work. |
|
I'm going to close this PR in favor of #9 |
Auto-generating them should reduce the potential for human error in
keeping them in sync.