Implement additional tags or a separate sink for process non-numerical data.

[zapdos26]

My team and I want to start collecting additional information around the processes for better troubleshooting. Info includes:

• user
• Specific environment variables
• cwd
• parent PID

We can accomplish this via the sysinfo library similar to the current host_metrics process. The problem we have (especially for the environment variables) is that the data is bulky. We are unsure if it would be advisable to add these as tags or make a new source specific for the non-numerical data around process.

Any guidance would be appreciated.

[pront]

Hi @zapdos26, your idea of collecting additional process-related information for troubleshooting makes sense, and the concern about data bulkiness is valid. There's also a privacy concern here, we don't want people's env vars to accidentally start leaking but this is relevant only if we go with a solution that touches existing components like host_metrics.

Some thoughts on the fly:

1. Add all these as new opt-in metrics to host_metrics
2. Introduce a new source for the environment variables. This could also accept an inclusion pattern to allow filtering. (This sounds might also benefit others users in the future)

• what about user, cwd, parent PID? They don't all fit in the new source. Maybe we could combine (1) and (2)

Let me know if this helps, and we can agree on a final direction.

[pront]

If you're still interested in a built-in source, please open a feature request with an example output record, target operating systems, and desired collection frequency so we can discuss the scope.