riakkv

Monitors one or more Riak KV servers.

Requirements:

The following charts are included, which are mostly derived from the metrics listed here.

  1. Throughput in operations/s
  • KV operations

    • gets
    • puts
  • Data type updates

    • counters
    • sets
    • maps
  • Search queries

    • queries
  • Search documents

    • indexed
  • Strong consistency operations

    • gets
    • puts
  1. Latency in milliseconds
  • KV latency of the past minute

    • get (mean, median, 95th / 99th / 100th percentile)
    • put (mean, median, 95th / 99th / 100th percentile)
  • Data type latency of the past minute

    • counter_merge (mean, median, 95th / 99th / 100th percentile)
    • set_merge (mean, median, 95th / 99th / 100th percentile)
    • map_merge (mean, median, 95th / 99th / 100th percentile)
  • Search latency of the past minute

    • query (median, min, max, 95th / 99th percentile)
    • index (median, min, max, 95th / 99th percentile)
  • Strong consistency latency of the past minute

    • get (mean, median, 95th / 99th / 100th percentile)
    • put (mean, median, 95th / 99th / 100th percentile)
  1. Erlang VM metrics
  • System counters

    • processes
  • Memory allocation in MB

    • processes.allocated
    • processes.used
  1. General load / health metrics
  • Siblings encountered in KV operations during the past minute

    • get (mean, median, 95th / 99th / 100th percentile)
  • Object size in KV operations during the past minute in KB

    • get (mean, median, 95th / 99th / 100th percentile)
  • Message queue length in unprocessed messages

    • vnodeq_size (mean, median, 95th / 99th / 100th percentile)
  • Index operations encountered by Search

    • errors
  • Protocol buffer connections

    • active
  • Repair operations coordinated by this node

    • read
  • Active finite state machines by kind

    • get
    • put
    • secondary_index
    • list_keys
  • Rejected finite state machines

    • get
    • put
  • Number of writes to Search failed due to bad data format by reason

    • bad_entry
    • extract_fail

configuration

The module needs to be passed the full URL to Riak’s stats endpoint. For example:

myriak:
  url: http://myriak.example.com:8098/stats

With no explicit configuration given, the module will attempt to connect to http://localhost:8098/stats.

The default update frequency for the plugin is set to 2 seconds as Riak internally updates the metrics every second. If we were to update the metrics every second, the resulting graph would contain odd jitter.