cLoki

like Loki, but for ClickHouse

cLoki is a flexible Loki ¹ & Tempo ¹ compatible LogQL API built on top of ClickHouse

Built in Explore UI and LogQL CLI for querying data
Native support Grafana ² and any LogQL clients for querying, processing, ingesting, tracing and alerting ³

Get started using the cLoki Wiki 💡

Project Background

The Loki API and its Grafana native integration are brilliant, simple and appealing - but we just love ClickHouse.

cLoki implements a complete LogQL API buffered by a fast bulking LRU sitting on top of ClickHouse tables and relying on its columnar search and insert performance alongside solid distribution and clustering capabilities for stored data. Just like Loki, cLoki does not parse or index incoming logs, but rather groups log streams using the same label system as Prometheus. ³

🔥 LogQL: Supported Features

cLoki implements a broad range of LogQL Queries to provide transparent compatibility with the Loki API
The Grafana Loki datasource can be used to natively query logs and display extracted timeseries

🎉 No plugins needed

🔥 Follow our examples to get started

⛽ Log Streams

cLoki supports input via Push API using JSON or Protobuf and it is compatible with Promtail and any other Loki compatible agent

Our preferred companion for parsing and shipping log streams to cLoki is paStash with extensive interpolation capabilities to create tags and trim any log fat. Sending JSON formatted logs is suggested when dealing with metrics.

🔥 CliQL: Experimental 2.0 Features

cLoki implements custom query functions for ClickHouse timeseries extraction, allowing direct access to any existing table

Timeseries

Convert columns to tagged timeseries using the emulated loki 2.0 query format

<aggr-op> by (<labels,>) (<function>(<metric>[range_in_seconds])) from <database>.<table> where <optional condition>

Examples

avg by (source_ip) (rate(mos[60])) from my_database.my_table

sum by (ruri_user, from_user) (rate(duration[300])) from my_database.my_table where duration > 10

ClickHouse

Convert columns to tagged timeseries using the experimental clickhouse function

Example

clickhouse({ 
  db="my_database", 
  table="my_table", 
  tag="source_ip", 
  metric="avg(mos)", 
  where="mos > 0", 
  interval="60" 
})

Query Options

parameter	description
db	clickhouse database name
table	clickhouse table name
tag	column(s) for tags, comma separated
metric	function for metric values
where	where condition (optional)
interval	interval in seconds (optional)
timefield	time/date field name (optional)

🔥 Tempo: Supported Features

cLoki Pulse offers experimental support for the Grafana Tempo API providing span ingestion and querying

At database level, Tempo Spans/Traces are stored as tagged Logs and are accessible from both LogQL and Tempo APIs

Setup

Check out the Wiki for detailed instructions or choose a quick method:

🚏 GIT (Manual)

Clone this repository, install with npmand run using nodejs 14.x (or higher)

npm install
CLICKHOUSE_SERVER="my.clickhouse.server" CLICKHOUSE_AUTH="default:password" CLICKHOUSE_DB="cloki" node cloki.js

🚏 NPM

Install cloki as global package on your system using npm

sudo npm install -g cloki
cd $(dirname $(readlink -f `which cloki`)) \
  && CLICKHOUSE_SERVER="my.clickhouse.server" CLICKHOUSE_AUTH="default:password" CLICKHOUSE_DB="cloki" cloki

🚏 PM2

sudo npm install -g cloki pm2
cd $(dirname $(readlink -f `which cloki`)) \
  && CLICKHOUSE_SERVER="my.clickhouse.server" CLICKHOUSE_AUTH="default:password" CLICKHOUSE_DB="cloki" pm2 start cloki
pm2 save
pm2 startup

🚏 Docker

For a fully working demo, check the docker-compose example

Logging

The project uses pino for logging and by default outputs JSON'ified log lines. If you want to see "pretty" log lines you can start cloki with npm run pretty

Configuration

The following ENV Variables can be used to control cLoki parameters and backend settings.

ENV	Default	Usage
CLICKHOUSE_SERVER	localhost	Clickhouse Server address
CLICKHOUSE_PORT	8123	Clickhouse Server port
CLICKHOUSE_DB	cloki	Clickhouse Database Name
CLICKHOUSE_AUTH	default:	Clickhouse Authentication (user:password)
CLICKHOUSE_PROTO	http	Clickhouse Protocol (http, https)
CLICKHOUSE_TIMEFIELD	record_datetime	Clickhouse DateTime column for native queries
BULK_MAXAGE	2000	Max Age for Bulk Inserts
BULK_MAXSIZE	5000	Max Size for Bulk Inserts
BULK_MAXCACHE	50000	Max Labels in Memory Cache
LABELS_DAYS	7	Max Days before Label rotation
SAMPLES_DAYS	7	Max Days before Timeseries rotation
HOST	0.0.0.0	cLOKi API IP
PORT	3100	cLOKi API PORT
CLOKI_LOGIN	undefined	Basic HTTP Username
CLOKI_PASSWORD	undefined	Basic HTTP Password
READONLY	false	Readonly Mode, no DB Init
FASTIFY_BODYLIMIT	5242880	API Maximum payload size in bytes
FASTIFY_REQUESTTIMEOUT	0	API Maximum Request Timeout in ms
FASTIFY_MAXREQUESTS	0	API Maximum Requests per socket
TEMPO_SPAN	24	Default span for Tempo queries in hours
TEMPO_TAGTRACE	false	Optional tagging of TraceID (expensive)
DEBUG	false	Debug Mode (for backwards compatibility)
LOG_LEVEL	info	Log Level
HASH	short-hash	Hash function using for fingerprints. Currently supported `short-hash` and `xxhash64` (xxhash64 function)

Disclaimer

cLoki is not affiliated or endorsed by Grafana Labs or ClickHouse Inc. All rights belong to their respective owners. ↩ ↩²
Grafana®, Loki™ and Tempo® are a Trademark of Raintank, Grafana Labs. ClickHouse® is a trademark of ClickHouse Inc. Prometheus is a trademark of The Linux Foundation. ↩
cLoki is a 100% clear-room api implementation and does not fork, use or derivate from Grafana Loki code or concepts. ↩ ↩²

kazu12/cLoki