InfluxDB 101

Connect to our instance

Curl

> curl -G 'https://scala-ci.typesafe.com/influx/query?pretty=true' --data-urlencode "u=scala" --data-urlencode "p=<>" --data-urlencode "db=scala_benchmark" --data-urlencode "q=SELECT * FROM result WHERE time > now() - 1d"
...

SSH and CLI

> ssh [email protected]
> influx -username scala -password <>
> use scala_benchmark
> SELECT * FROM result WHERE time > now() - 1d
...

Web UI

> ssh [email protected] -L 8086:localhost:8086 -N &
> ssh [email protected] -L 8083:localhost:8083 -N &

Open http://localhost:8083. Host: localhost, port: 8086, no SSL.

In principle, it should work without the first tunnel (host: scala-ci.typesafe.com/influx, port: 443, with SSL), but I couldn't get it to work.

Try with query SELECT * FROM result WHERE time > now() - 1d.

InfluxDB schema

Measurement

• Similar to an sql table
• Examples in scala_benchmark: result, commit
• Query: SHOW MEASUREMENTS

Tags

• Similar to an indexed column in sql
• Examples: benchmark, branch, source, ...
• Use for commonly queried meta-data
• Tag values are always strings
• Queries: SHOW TAG KEYS FROM result, SHOW TAG VALUES FROM result WITH KEY = benchmark

Fields

• Similar to non-indexed columns in sql
• Data and metadata
• Examples: javaVersion, score, ...
• Query: SHOW FIELD KEYS FROM result

Point

• A data entry, similar to a row in sql
• Contains a timestamp, values for all tag keys, values for some field keys (sparse)

Series

• A measurement with a unique "tag set" (a set of tag key-value pairs)
• Example: result,benchmark=HotScalacBenchmark.compile,branch=2.12.x,source=better-files

Writing the same (measurement,timestamp,tagSet) twice overwrites the existing point. We can exploit this to re-run a benchmark.

Working with data

Querying

https://docs.influxdata.com/influxdb/v1.2/query_language/data_exploration/

• A query requires at least one field key in the SELECT clause to return data
• Tag and field keys can be quoted in double quotes (if necessary)
• String literals need to be in single quotes
  • Example: SELECT * FROM result WHERE "benchmark" = 'HotScalacBenchmark.compile' LIMIT 5
• Regex matching: WHERE tagName =~ /regex/
• Query by time: SELECT * FROM result WHERE time > now() - 7d
  • now() is the current time in nanoseconds

Entering data

InfluxDB is a schemaless database: measurements, tags, and fields are created on demand.

Write commands

• https://docs.influxdata.com/influxdb/v1.2/write_protocols/line_protocol_reference/
• Example: 'result,benchmark=<>,branch=<>... javaVersion=<>,score=<>,... 1434055562000000000'

Types

• Measurements, tag keys, tag values and field keys are strings. Don't double-quote!
  • If a tag is not provided, its value is ''. If a write introduces a new tag, all existing points get tag value ''.
• Timestamps are unix timestamps in nanos (365199550781253)
• Types for field values:
  • 1, 1.0 are Double
  • 1i is Int
  • t, F, true, False, TRUE are Boolean
  • "s" is String

Writing with curl

curl -i -XPOST 'https://scala-ci.typesafe.com/influx/write?db=scala_benchmark' --data-binary 'result,benchmark=<>,branch=<>... javaVersion=<>,score=<>,... 1434055562000000000'

Back up / copy a measurement

SELECT * INTO new_name FROM old_name GROUP BY *

• GROUP BY * is important to preserve tags as tags, otherwise they are converted to fields which causes data loss.
• http://docs.influxdata.com/influxdb/v1.2/query_language/data_exploration/#common-issues-with-the-into-clause
• http://docs.influxdata.com/influxdb/v1.2/troubleshooting/frequently-asked-questions/#why-are-my-into-queries-missing-data
• https://github.com/influxdata/influxdb/issues/4155#issuecomment-268719194

Deleting data

Delete one or multiple series from a measurement: DROP SERIES

• Can be used to drop a tag that's no longer in use

Delete an entire measurement: DROP MEASUREMENT

Fields cannot be deleted, need to copy into a new measurement

• http://stackoverflow.com/questions/26767148/how-to-delete-columns-of-a-series-in-influxdb/