Allow specifying hostname and ip via tags (#167)

* Allow specifying the hostname via a tag

This is useful for applications running inside containers where the
hostname is autogenerated and generally useless. In kubernetes, the
downward api can be used to pass in the node name actually running
the container.

Fixes #66

Signed-off-by: Jeff Schroeder <[email protected]>

* Catch the `socket.error` exception when trying to get the hostname

This is the base exception for all errors in the socket library. Explicit
is better than implicit!

Signed-off-by: Jeff Schroeder <[email protected]>

* Don't attempt to lookup the ip address if specified via a tag

This matches the functionality of jaeger-client-java after
jaegertracing/jaeger-client-java#371 was merged.

Refs: #166

Signed-off-by: Jeff Schroeder <[email protected]>

* Remove ipv4_to_int since jaeger converts tags to strings

Signed-off-by: Jeff Schroeder <[email protected]>

Author: SEJeff

jaeger_client/thrift.py

@@ -13,8 +13,6 @@
 # limitations under the License.
 
 import six
-import socket
-import struct
 
 import jaeger_client.thrift_gen.jaeger.ttypes as ttypes
 import jaeger_client.thrift_gen.sampling.SamplingManager as sampling_manager
@@ -28,17 +26,6 @@
     long = int
 
 
-def ipv4_to_int(ipv4):
-    if ipv4 == 'localhost':
-        ipv4 = '127.0.0.1'
-    elif ipv4 == '::1':
-        ipv4 = '127.0.0.1'
-    try:
-        return struct.unpack('!i', socket.inet_aton(ipv4))[0]
-    except:
-        return 0
-
-
 def id_to_int(big_id):
     if big_id is None:
         return None

jaeger_client/tracer.py

@@ -29,7 +29,6 @@
 from .codecs import TextCodec, ZipkinCodec, ZipkinSpanFormat, BinaryCodec
 from .span import Span, SAMPLED_FLAG, DEBUG_FLAG
 from .span_context import SpanContext
-from .thrift import ipv4_to_int
 from .metrics import Metrics, LegacyMetricsFactory
 from .utils import local_ip
 
@@ -53,7 +52,6 @@ def __init__(
         self.service_name = service_name
         self.reporter = reporter
         self.sampler = sampler
-        self.ip_address = ipv4_to_int(local_ip())
         self.metrics_factory = metrics_factory or LegacyMetricsFactory(metrics or Metrics())
         self.metrics = TracerMetrics(self.metrics_factory)
         self.random = random.Random(time.time() * (os.getpid() or 1))
@@ -80,16 +78,19 @@ def __init__(
             self.codecs.update(extra_codecs)
         self.tags = {
             constants.JAEGER_VERSION_TAG_KEY: constants.JAEGER_CLIENT_VERSION,
-            constants.JAEGER_IP_TAG_KEY: self.ip_address,
         }
         if tags:
             self.tags.update(tags)
-        # noinspection PyBroadException
-        try:
-            hostname = socket.gethostname()
-            self.tags[constants.JAEGER_HOSTNAME_TAG_KEY] = hostname
-        except:
-            logger.exception('Unable to determine host name')
+
+        if self.tags.get(constants.JAEGER_IP_TAG_KEY) is None:
+            self.tags[constants.JAEGER_IP_TAG_KEY] = local_ip()
+
+        if self.tags.get(constants.JAEGER_HOSTNAME_TAG_KEY) is None:
+            try:
+                hostname = socket.gethostname()
+                self.tags[constants.JAEGER_HOSTNAME_TAG_KEY] = hostname
+            except socket.error:
+                logger.exception('Unable to determine host name')
 
         self.reporter.set_process(
             service_name=self.service_name,

tests/test_thrift.py

@@ -24,13 +24,6 @@
 from thrift.transport.TTransport import TMemoryBuffer
 
 
-def test_ipv4_to_int():
-    base = thrift.ipv4_to_int('127.0.0.1')
-    assert thrift.ipv4_to_int('localhost') == base
-    assert thrift.ipv4_to_int('::1') == base
-    assert thrift.ipv4_to_int('a:b:1') == 0
-
-
 def test_submit_batch(tracer):
     span = tracer.start_span("test-span")
     span.set_tag('bender', 'is great')

tests/test_tracer.py

@@ -15,6 +15,7 @@
 import mock
 import random
 import six
+import socket
 
 import pytest
 import tornado.httputil
@@ -200,7 +201,7 @@ def test_tracer_tags_no_hostname():
     from jaeger_client.tracer import logger
     with mock.patch.object(logger, 'exception') as mock_log:
         with mock.patch('socket.gethostname',
-                        side_effect=['host', ValueError()]):
+                        side_effect=['host', socket.timeout()]):
             Tracer(service_name='x', reporter=reporter, sampler=sampler)
         assert mock_log.call_count == 1
 
@@ -260,3 +261,31 @@ def test_tracer_override_codecs():
                 "Extra codec not found"
         assert tracer.codecs[Format.BINARY] == "overridden_binary_codec",\
                 "Binary format codec not overridden"
+
+
+def test_tracer_hostname_tag():
+    reporter = mock.MagicMock()
+    sampler = ConstSampler(True)
+    tracer = Tracer(
+        service_name='x',
+        tags={c.JAEGER_HOSTNAME_TAG_KEY: 'jaeger-client-app.local'},
+        reporter=reporter,
+        sampler=sampler,
+    )
+
+    assert tracer.tags[c.JAEGER_HOSTNAME_TAG_KEY] == \
+            'jaeger-client-app.local'
+
+
+def test_tracer_ip_tag():
+    reporter = mock.MagicMock()
+    sampler = ConstSampler(True)
+    tracer = Tracer(
+        service_name='x',
+        tags={c.JAEGER_IP_TAG_KEY: '192.0.2.3'},
+        reporter=reporter,
+        sampler=sampler,
+    )
+
+    assert tracer.tags[c.JAEGER_IP_TAG_KEY] == \
+            '192.0.2.3'