ui: implementation of transient stats for admin UI

[spencerkimball]

Create new statistics for transient stats at:

• Gateways for client traffic

  • of requests

  • of bytes in

  • of bytes out

  • of failures

  • Latency of client requests
• Node level (both for Raft & KV traffic)

  • of RPCs

  • of bytes in

  • of bytes out

  • of failed RPCs

  • Latency of internode requests
• Store level

  • of store requests

  • of bytes to storage

  • of bytes from storage

  • of failed commands

  • Latency of store requests

[tbg]

partial server-side work in #3504 should make it easy to access some of that information already. Others can be added easily. @maxlang, let's talk when you get back.

SQL is missing from the above list.

[maxlang]

Here's the new list we're discussing supporting

Primary (empty)

• Status - gr health, y unhealthy, r dead ( counts/status): 20 / 4 / 1
• Node name (# nodes)
• Stores (total stores)
• Ranges (total ranges)

Storage

• total count/bytes - 180 (54 KiB)
• Live count/bytes - (live count / live value bytes + key bytes)
• intent count/bytes
• sys count/bytes - metadata, trans records, etc

Gateway Traffic - these are totals, or viewable by time range

• SQL CRUD stats
  • TRANSACTIONS (count)
  • SELECTS (count, bytes out)
  • UPDATES (count, bytes in)
  • INSERTS (count, bytes in)
  • DELETES (count)
  • CREATES (count)
  • DROPS (count)

Internal traffic

• Total reqs count/bytes in /bytes out
• Req Reads count/ bytes out
• Req Writes count / bytes in

Range Activity

• Leaders (count)
• Splits
• Merges
• Rebalances
• Availability (%)

[cuongdo]

Also need ranges / table endpoint

[spencerkimball]

@cuongdo should this be closed?

[cuongdo]

Not yet. I'm dividing this up into three pieces for manageability.

1. SQL (done)
2. Transactions (SQL & KV) -- in-progress
3. Internal traffic - to be done after grpc stuff settles down

[cuongdo]

SQL & KV are now done.

We're punting on on the internal traffic stats until after 1.0. I'll create a new issue for that.