Hello fellow Aurora enthusiasts,
I am struggling with the docker containerizer. Jobs launched using docker
containers end up failing, because Thermos cannot find Mesos's required
dependencies, such as libcurl, libsasl, libsvn… See trace below, with libsasl2
:
Traceback (most recent call last):
File "apache/aurora/executor/bin/thermos_executor_main.py", line 45, in
<module>
from mesos.native import MesosExecutorDriver
File
"/root/.pex/install/mesos.native-0.27.2-py2.7-linux-x86_64.egg.6c53d177e078864264dbb5af282e84d7c2cdc3a8/mesos.native-0.27.2-py2.7-linux-x86_64.egg/mesos/native/__init__.py",
line 17, in <module>
from ._mesos import MesosExecutorDriverImpl
ImportError: libsasl2.so.2: cannot open shared object file: No such file or
directory
Having stumbled upon the issue quite a lot, I found a way to fix it : having
the required dependencies installed on the containers I need to run. It’s
annoying and feels kinda dirty though. I dug into Mesos documentation and found
this at http://mesos.apache.org/documentation/latest/container-image/ :
For a general purpose executor (e.g., thermos) of a framework (e.g., Aurora),
requiring it and all its dependencies to be present in all possible container
images that a user might use is not trivial.
Therefore, my question is : do I need to add Mesos’s dependencies to each and
every docker image I want to run (and is this best practice) or am I doing
something terribly wrong ?
Thanks for your support !
Adrian
