zipkin.storage

Interface StorageComponent.Builder

All Known Implementing Classes:

InMemoryStorage.Builder

Enclosing interface:

StorageComponent

public static interface StorageComponent.Builder

Method Summary

Modifier and Type	Method and Description
StorageComponent	build()
StorageComponent.Builder	strictTraceId(boolean strictTraceId) Zipkin supports 64 and 128-bit trace identifiers, typically serialized as 16 or 32 character hex strings.

Method Detail

strictTraceId

StorageComponent.Builder strictTraceId(boolean strictTraceId)

Zipkin supports 64 and 128-bit trace identifiers, typically serialized as 16 or 32 character hex strings. When false, this setting only considers the low 64-bits (right-most 16 characters) of a trace ID when grouping or retrieving traces. This should be set to false while some applications issue 128-bit trace IDs and while other truncate them to 64-bit. If 128-bit trace IDs are not in use, this setting is not required.

Details

Zipkin historically had 64-bit trace IDs, but it now supports 128-bit trace IDs via Span.traceIdHigh, or its 32-character hex representation. While instrumentation update to propagate 128-bit IDs, it can be ambiguous whether a 64-bit trace ID was sent intentionally, or as an accident of truncation. This setting allows Zipkin to be usable until application instrumentation are upgraded to support 128-bit trace IDs.

Here are a few trace IDs the help explain this setting.

   
Trace ID A: 463ac35c9f6413ad48485a3953bb6124
   
Trace ID B: 48485a3953bb6124
   
Trace ID C: 463ac35c9f6413adf1a48a8cff464e0e
   
Trace ID D: 463ac35c9f6413ad
 

In the above example, Trace ID A and Trace ID B might mean they are in the same trace, since the lower-64 bits of the IDs are the same. This could happen if a server A created the trace and propagated it to server B which ran an older tracing library. Server B could have truncated the trace ID to lower-64 bits. When strictTraceId == false, spans matching either trace ID A or B would be returned in the same trace when searching by ID A or B. Spans with trace ID C or D wouldn't be when searching by ID A or B because trace IDs C and D don't share lower 64-bits (right-most 16 characters) with trace IDs A or B.

It is also possible that all servers are capable of handling 128-bit trace identifiers, but are configured to only send 64-bit ones. In this case, if strictTraceId == false trace ID A and B would clash and be put into the same trace, causing confusion. Moreover, there is overhead associated with indexing spans both by 64 and 128-bit trace IDs. When a site has finished upgrading to 128-bit trace IDs, they should enable this setting.

See https://github.com/openzipkin/b3-propagation/issues/6 for the status of known open source libraries on 128-bit trace identifiers.

build

StorageComponent build()