I ran into this issue as well - it looks like the problem is that Mesos requires protobuf<3, which 3.0.0-alpha1 technically is (pants uses a different resolver than pip - pip will ignore alpha versions unless run with --pre). A workaround is to add an explicit dependency on protobuf==2.6.1 to our 3rdparty/python/requirements.txt and add a dependency on 3rdpary/python:protobuf everywhere that depends on 3rdparty/python:mesos.native.
If you fix this, it'd be awesome if you could submit a patch. Thanks, Kevin On Tue, Feb 17, 2015 at 1:06 AM, Erb, Stephan <stephan....@blue-yonder.com> wrote: > Hi everyone, > > we try to build the latest Aurora git version on a Debian 7 host. > Unfortunately, the resulting executor seems to be broken, as it fails to > start with the following message: > > > WARNING: Logging before InitGoogleLogging() is written to STDERR > I0216 18:16:51.178513 25237 fetcher.cpp:76] Fetching URI > '/opt/thermos/bin/thermos_executor.sh' > I0216 18:16:51.178513 25237 fetcher.cpp:179] Copying resource from > '/opt/thermos/bin/thermos_executor.sh' to > '/var/lib/mesos/slaves/20141218-095422-1985259712-5050-31688-0/frameworks/20140919-174559-16842879-5050-27194-0000/executors/thermos-1424107010619-stratosphere-staging-strato-stage-service-71-0-9120fb5c-4b4b-425e-90f5-0d83224b9d64/runs/fe198e72-d41e-41c3-a006-27dad0ed758b' > Traceback (most recent call last): > File "/opt/thermos/bin/thermos_executor.pex/.bootstrap/_pex/pex.py", > line 272, in execute > self.execute_entry(entry_point, args) > File "/opt/thermos/bin/thermos_executor.pex/.bootstrap/_pex/pex.py", > line 320, in execute_entry > runner(entry_point) > File "/opt/thermos/bin/thermos_executor.pex/.bootstrap/_pex/pex.py", > line 342, in execute_pkg_resources > runner = entry.load(require=False) # trust that the environment is > sane > File > "/opt/thermos/bin/thermos_executor.pex/.bootstrap/pkg_resources.py", line > 2048, in load > entry = __import__(self.module_name, globals(),globals(), ['__name__']) > File > "/opt/thermos/bin/thermos_executor.pex/apache/aurora/executor/bin/thermos_executor_main.py", > line 24, in <module> > File > "/.pex/install/mesos.native-0.21.1-py2.7-linux-x86_64.egg.b3704775c8aad2e57af36756ac89f16a8ca72ae1/mesos.native-0.21.1-py2.7-linux-x86_64.egg/mesos/native/__init__.py", > line 17, in <module> > from ._mesos import MesosExecutorDriverImpl > File > "/.pex/install/mesos.interface-0.21.1-py2-none-any.whl.00a9556e7f26cf49f09f1542a1ee274c058ea78c/mesos.interface-0.21.1-py2-none-any.whl/mesos/interface/mesos_pb2.py", > line 4, in <module> > from google.protobuf.internal import enum_type_wrapper > ImportError: cannot import name enum_type_wrapper > > > I have attached the pants build log. It seems like pants is pulling in a > wrong protobuf version (protobuf-3.0.0_alpha_1 instead of protobuf 2.6.1). > > The issue can even be simulated locally: > * pip install mesos.interface with an old virtualenv 1.11 (including pip > 1.5 and setuptools 2.0.2) pulls in protobuf-3.0.0-alpha-1 > * pip install mesos.interface with the latest and greatest pip 6.0.8 and > setuptools 12.2 pulls protobuf-2.6.1 > > How can this happen? Isn't the "build-support/virtualenv" script supposed > to prevent that? I also tried patching the setuptools_version in pants.ini, > but it did not help. > > Any ideas? > > Best Regards, > Stephan > > > > >
