ClickHouse/utils/trace-visualizer

Trace visualizer is a tool for representation of a tracing data as a Gantt diagram.

Quick start

For now this tool is not integrate into clickhouse and require a lot of manual adjustments.

cd utils/trace-visualizer
python3 -m http.server

Open localhost. It will show an example of data. To show your tracing data you have to put it in JSON format near index.html and change call to fetchData() function at the bottom of index.html. (Do not forget to disable browser caching while changing it).

Visualizing query trace

First of all opentelemetry_span_log system table must be enabled to save query traces. Then run a query you want to trace with a setting:

set opentelemetry_start_trace_probability=1;
SELECT 1;

To find out trace_id of a query run the following command:

SELECT DISTINCT trace_id FROM system.opentelemetry_span_log ORDER BY query_start_time DESC;

To obtain JSON data suitable for visualizing run:

SELECT tuple (parent_span_id, attribute['clickhouse.thread_id'] || attribute['thread_number'] as thread_id)::Tuple(parent_span_id UInt64, thread_id String) as group, operation_name, start_time_us, finish_time_us, sipHash64(operation_name) as color, attribute
from system.opentelemetry_span_log
WHERE trace_id = 'your-trace-id'
ORDER BY group ASC
FORMAT JSON SETTINGS output_format_json_named_tuples_as_objects = 1;

Dependencies

1. D3js (v4).
2. Tooltips for D3.
3. jquery.
4. Bootstrap.