tian-junwei / zipkin-storage-postgresql

storage-postgresql

This PostgreSQL storage component includes a blocking SpanStore and span consumer function. SpanStore.getDependencies() aggregates dependency links on-demand.

The implementation uses JOOQ to generate PostgreSQL SQL commands. It is only tested on postgresql 9.5.

See the schema DDL.

zipkin.storage.postgresql.PostgreSQLStorage.Builder includes defaults that will operate against a given Datasource.

Testing this component

This module conditionally runs integration tests against a local PostgreSQL instance.

You minimally need to export the variable POSTGRESQL_USER to run tests. Ex.

$ POSTGRESQL_USER=postgres ./mvnw clean install -pl :zipkin-storage-mysql

If you run tests via Maven or otherwise without specifying POSTGRESQL_USER, you'll notice tests are silently skipped.

Results :

Tests run: 49, Failures: 0, Errors: 0, Skipped: 48

This behaviour is intentional: We don't want to burden developers with installing and running all storage options to test unrelated change. That said, all integration tests run on pull request via Travis.

Exploring Zipkin Data

When troubleshooting, it is important to note that zipkin ids are encoded as hex. If you want to view data in mysql, you'll need to use the hex function accordingly.

For example, all the below query the same trace using different tools:

• zipkin-ui: http://1.2.3.4:9411/traces/27960dafb1ea7454
• zipkin-api: http://1.2.3.4:9411/api/v1/trace/27960dafb1ea7454?raw
• postgresql: select * from zipkin_spans where trace_id = x'27960dafb1ea7454';

If you are trying to debug from data in the database, it is helpful to format IDs as hex, and timestamps as dates. The following is an example query which will return one line for each update to a span in the last 5 minutes.

SELECT lower(concat(CASE trace_id_high
                        WHEN '0' THEN ''
                        ELSE hex(trace_id_high)
                    END,hex(trace_id))) AS trace_id,
       lower(hex(parent_id)) as parent_id,
       lower(hex(id)) as span_id,
       name,
       from_unixtime(start_ts/1000000) as timestamp
FROM zipkin_spans
where (start_ts/1000000) > UNIX_TIMESTAMP(now()) - 5 * 60;

For example, the output below shows two traces recently reported. One of which is using 128-bit trace IDs. You could copy and paste the trace_id into zipkin's UI to troubleshoot further.

+----------------------------------+------------------+------------------+------+--------------------------+
| trace_id                         | parent_id        | span_id          | name | timestamp                |
+----------------------------------+------------------+------------------+------+--------------------------+
| abbd9f5da49e5848aa4b729ff2bc90a3 | NULL             | aa4b729ff2bc90a3 | get  | 2017-04-19 12:43:00.9830 |
| abbd9f5da49e5848aa4b729ff2bc90a3 | aa4b729ff2bc90a3 | 7888a4aef81f074d | get  | 2017-04-19 12:43:01.1960 |
| 11b98d7107dac980                 | 11b98d7107dac980 | bc33c2d5ad25bf89 | get  | 2017-04-19 12:42:45.4240 |
| 11b98d7107dac980                 | NULL             | 11b98d7107dac980 | get  | 2017-04-19 12:42:45.0680 |
+----------------------------------+------------------+------------------+------+--------------------------+

Applying the schema

execute schema DDL. into PostgreSQL Server

Generating the schema types

$ rm -rf zipkin-storage/postgresql/src/main/java/zipkin/storage/postgresql/internal/generated/
$ ./mvnw -pl :zipkin-storage-postgresql clean org.jooq:jooq-codegen-maven:generate com.mycila:license-maven-plugin:format