Skip to content
  • v1.16.0
  • 1c31578
  • Compare
    Choose a tag to compare
    Search for a tag
  • v1.16.0
  • 1c31578
  • Compare
    Choose a tag to compare
    Search for a tag

@pavolloffay pavolloffay released this Dec 17, 2019

Backend Changes

Breaking Changes

List of service operations can be classified by span kinds (#1943, #1942, #1937, @guo0693)
  • Endpoint changes:
    • Both Http & gRPC servers now take new optional parameter spanKind in addition to service. When spanKind
      is absent or empty, operations from all kinds of spans will be returned.
    • Instead of returning a list of string, both Http & gRPC servers return a list of operation struct. Please
      update your client code to process the new response. Example response:
      curl 'http://localhost:6686/api/operations?service=UserService&spanKind=server' | jq
      {
          "data": [{
              "name": "UserService::getExtendedUser",
              "spanKind": "server"
          },
          {
              "name": "UserService::getUserProfile",
              "spanKind": "server"
          }],
          "total": 2,
          "limit": 0,
          "offset": 0,
          "errors": null
      }
      
    • The legacy http endpoint stay untouched:
      /services/{%s}/operations
      
  • Storage plugin changes:
    • Memory updated to support spanKind on write & read, no migration is required.
    • Badger & ElasticSearch
      to be implemented:
      For now spanKind will be set as empty string during read & write, only name will be valid operation name.
    • Cassandra updated to support spanKind on write & read (#1937, @guo0693):
      If you don't run the migration script, nothing will break, the system will used the old table
      operation_names and set empty spanKind in the response.
      Steps to get the updated functionality:
      1. You will need to run below command on the host you can use cqlsh to connect the the cassandra contact
        point
        KEYSPACE=jaeger_v1 CQL_CMD='cqlsh host 9042 -u test_user -p test_password --request-timeout=3000' 
        bash ./v002tov003.sh
        
        The script will create new table operation_names_v2 and migrate data from the old table.
        spanKind column will be empty for those data.
        At the end, it will ask you whether you want to drop the old table or not.
      2. Restart ingester & query services so that they begin to use the new table
Trace and Span IDs are always padded to 32 or 16 hex characters with leading zeros (#1956, @yurishkuro)

Previously, Jaeger backend always rendered trace and span IDs as the shortest possible hex string, e.g. an ID
with numeric value 255 would be rendered as a string ff. This change makes the IDs to always render as 16 or 32
characters long hex string, e.g. the same id=255 would render as 00000000000000ff. It mostly affects how UI
displays the IDs, the URLs, and the JSON returned from jaeger-query service.

Motivation: Among randomly generated and uniformly distributed trace IDs, only 1/16th of them start with 0
followed by a significant digit, 1/256th start with two 0s, and so on in decreasing geometric progression.
Therefore, trimming the leading 0s is a very modest optimization on the size of the data being transmitted or stored.

However, trimming 0s leads to ambiguities when the IDs are used as correlations with other monitoring systems,
such as logging, that treat the IDs as opaque strings and cannot establish the equivalence between padded and
unpadded IDs. It is also incompatible with W3C Trace Context and Zipkin B3 formats, both of which include all
leading 0s, so an application instrumented with OpenTelemetry SDKs may be logging different trace ID strings
than application instrumented with Jaeger SDKs (related issue #1657).

Overall, the change is backward compatible:

  • links with non-padded IDs in the UI will still work
  • data stored in Elasticsearch (where IDs are represented as strings) is still readable

However, some custom integration that rely on exact string matches of trace IDs may be broken.

Change default rollover conditions to 2 days (#1963, @pavolloffay)

Change default rollover conditions from 7 days to 2 days.

Given that by default Jaeger uses daily indices and some organizations do not keep data longer than 7 days
the default of 7 days seems unreasonable - it might result in a too big index and
running curator would immediately remove the old index.

New Features

Bug fixes, Minor Improvements

UI Changes

  • UI pinned to version 1.6.0. The changelog is available here v1.6.0
Assets 5
You can’t perform that action at this time.