Make prometheus-dirsize-reporter more robust

[consideRatio]

https://github.com/yuvipanda/prometheus-dirsize-exporter is used for our shared-dirsize-metrics k8s Deployment that ships with each JupyterHub. It is currently crashing for some hubs (nasa-veda for example), and due to that never enters a ready state. This in turn makes our CI/CD pipeline fail when making upgrades to hubs as help upgrade waits for its deployed resources succesful startup.

This issue is meant to help us track and get various upstream work done in this project to help us resolve the short term need, but also to ensure the projects meets some key critiera we would appreciate seeing in any upstream project we rely on. Below I've picked out some issues that I think we should resolve.

Immediate needs

• Missing filesystem permissions can cause a crash loop yuvipanda/prometheus-dirsize-exporter#5 and "Stale file handle" can cause crash loop yuvipanda/prometheus-dirsize-exporter#6
  • Addressed by - Handle permission errors and stale file errors better yuvipanda/prometheus-dirsize-exporter#7

Other needs

• Add tests yuvipanda/prometheus-dirsize-exporter#9
• Followup on error handling to avoid discarding too much information yuvipanda/prometheus-dirsize-exporter#10
• Improved logging - startup info, timestamps, etc yuvipanda/prometheus-dirsize-exporter#2
• Add CHANGELOG.md and RELEASE.md yuvipanda/prometheus-dirsize-exporter#11

[yuvipanda]

Thanks for opening this, @consideRatio and listing the various things that need to be done. I've just retitled the issue to sound a little gentler, hope that's ok.

[consideRatio]

Thank you @yuvipanda, it also reads more clearly!

[yuvipanda]

One round of robustness complete, let's close this.