py_zipkin span is not accepted by the openzipkin collector.

1. Run docker run -d --net host openzipkin/zipkin

2. Install py-zipkin pip install py-zipkin==0.8.3

3. Run the following python script:

#!/usr/bin/env python

import requests

from py_zipkin.zipkin import zipkin_span

def http_transport(encoded_span):
    # The collector expects a thrift-encoded list of spans.
    requests.post(
        'http://localhost:9411/api/v1/spans',
        data=encoded_span,
        headers={'Content-Type': 'application/x-thrift'},
    )

if __name__ == '__main__':
    with zipkin_span(
        service_name='test_service',
        transport_handler=http_transport,
        sample_rate=100.0,
    ):
        print "Work"

I see the following log line in the collector and get a 400 responce

2017-08-07 00:43:04.851  WARN 7 --- [nio-9411-exec-7] zipkin.server.ZipkinHttpCollector        : [Malformed reading List<Span> from TBinary]([url](url))

java.lang.IllegalArgumentException: Malformed reading List<Span> from TBinary
	at zipkin.internal.ThriftCodec.exceptionReading(ThriftCodec.java:600) ~[zipkin-1.29.1.jar!/:na]
	at zipkin.internal.ThriftCodec.readList(ThriftCodec.java:564) ~[zipkin-1.29.1.jar!/:na]
	at zipkin.internal.ThriftCodec.readSpans(ThriftCodec.java:88) ~[zipkin-1.29.1.jar!/:na]
	at zipkin.collector.Collector.acceptSpans(Collector.java:93) ~[zipkin-1.29.1.jar!/:na]
	at zipkin.server.ZipkinHttpCollector.validateAndStoreSpans(ZipkinHttpCollector.java:88) [classes/:na]
	at zipkin.server.ZipkinHttpCollector.uploadSpansThrift(ZipkinHttpCollector.java:73) [classes/:na]

[cybergrind]

I believe the core difference is the https://github.com/Yelp/py_zipkin/blob/master/py_zipkin/thrift/__init__.py#L181

My issue was caused by using old http transport code, that basically was:

def http_transport(encoded_span):
    # copypasted from: https://github.com/Yelp/py_zipkin#transport
    # The collector expects a thrift-encoded list of spans. Instead of
    # decoding and re-encoding the already thrift-encoded message, we can just
    # add header bytes that specify that what follows is a list of length 1.
    body = b'\x0c\x00\x00\x00\x01' + encoded_span
    requests.post(
        'http://{}/api/v1/spans'.format(ZIPKIN),
        data=body,
        headers={'Content-Type': 'application/x-thrift'},
    )

As I understand: b’\x0c\x00\x00\x00\x01’ - stands for list with length 1 and 0.9.0 already encodes everything as a list so it doesn’t required anymore.

[bplotnick]

@cybergrind Thanks for posting this. We should’ve been clearer about that backwards incompatibility. Glad you got it resolved.

I’m going to go ahead and close this. Others can open a new issue or reach out on the zipkin gitter if they run into similar problems.