Improving guidelines to reaching 200k metric samples/s

[JamesGuthrie]

The following are my rough notes on things that we can do to improve the user experience of reaching 200k metric samples/s:

• max_wal_size
  • We can provide an actual config recommendation based on samples/sec
    • (possibly also) Tools to determine how to know how big the ideal WAL size is
  • Add recommendation in runbook about checking the checkpoint req count when metric ingest is lower than expected
• wal_compression: should be on by default
• Tuning chunk index size / chunk interval
  • Examining active chunk index size vs. shared buffers
  • Configuring default chunk interval
  • Configuring per-metric chunk interval
• Disk recommendation:
  • We don't have a minimum required disk iops/throughput for achieving 200k samples/sec