----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/41869/#review112502 ----------------------------------------------------------- Thanks Tim for taking care of this. I have been thinking that we should have TokenManager keep a map of -> . This would help us to have any realm being handled by the registry_client + token manager. This way, all pull requests will request the TokenManager for a token with ```getToken(realm, service, scope)```. TokenManager will lookup the cache and return the token or create a new item in the cache (after getting the token from realm). What do you think? - Jojy Varghese On Jan. 4, 2016, 2:13 a.m., Timothy Chen wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/41869/ > ----------------------------------------------------------- > > (Updated Jan. 4, 2016, 2:13 a.m.) > > > Review request for mesos, Gilbert Song, Jie Yu, and Jojy Varghese. > > > Repository: mesos > > > Description > ------- > > Currently Docker auth server is being configured via docker_auth_server flag, but we should actually get the docker auth server information from docker registry. This allows us to remove one flag and also support both public and private docker registry. > > > Diffs > ----- > > src/slave/containerizer/mesos/provisioner/docker/registry_client.hpp e73d4d785bccf1d50af9e91d2d20dafb444bec68 > src/slave/containerizer/mesos/provisioner/docker/registry_client.cpp 4e305ade1f9f8b42de5c8db636c6d3f5d8e2444f > src/slave/containerizer/mesos/provisioner/docker/registry_puller.cpp 049d46cad5cf94a3fb5d74cbfe649850311d35ad > src/slave/flags.hpp 2b2679c1ae68d120756eaf81e5728d20791d6746 > src/slave/flags.cpp a60d3c8022aba93fbd17a46dfff601fb1b25bbee > src/tests/containerizer/provisioner_docker_tests.cpp d51f342dabf386fb618ef72ce3e36a8bd8c82b5f > > Diff: https://reviews.apache.org/r/41869/diff/ > > > Testing > ------- > > > Thanks, > > Timothy Chen > >