Beats diagnostics tooling

[urso]

Beats should provide a tool that allows one to easily collect information that can help to analyze known user issues. Diagnostics should contain static information like host meta data, configuration files and existing log files. Active data collection should also be supported, like metrics or memory profiles.

Tasks Phase 1:

• Enable http endpoint with unix socket or named pipe by default. e.g. /var/lib/<meta-hash>.sock - completed as part of previous work to enable monitoring
• Add API to read current Beats configuration. For now the static configuration used on startup, including modules. - completed as part of previous work for elastic-agent inspect
• Introduce CLI subcommand 'diag collect' collecting diagnostics into a zip file:
  • query Beat info via unix socket
  • query active Beat configuration
• Collect the Beat instance log files if file output has been configured.
• Beats diagnostic tooling - Phase II elastic-agent#94
  • Libbeat collects metrics every 10s into ring buffer (N=60)
  • Expose internal metrics via API
  • Diagnostics tool reads metrics ring buffer
• (optional) CLI subcommand diag add to add extra files and directories to a diagnostics report.
• log / capture std from the processes started

[elasticmachine]

Pinging @elastic/integrations-services (Team:Services)

[jlind23]

@michel-laterman should that one also be part of 8.0 or everything has been done in 7.16?

[michel-laterman]

The diagnostics collect command is being backported to 7.16, it can gather meta-data, configuration, the active policy, and local logs.

[jlind23]

@michel-laterman just added a task around: log / capture std from the processes started

[jlind23]

Closing - Future work will be track there: elastic/elastic-agent#94