[
https://issues.apache.org/jira/browse/MESOS-6909?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15882272#comment-15882272
]
Milan Baran commented on MESOS-6909:
------------------------------------
Well, I have done some further investigation. It looks like *'path -H socket ps
-a'* can run without problem. The difference between *ps* and *pull* is that
*ps* hasn't set environments argument and as it says in codes
_mesos/3rdparty/libprocess/include/process/subprocess_base.hpp_
{code}
* @param environment Environment variables to use for the new
* subprocess or if None (the default) then the new subprocess
* will inherit the environment of the current process.
{code}
Could we consider to make more robust with introducing a
MESOS_INHERITANCE_STRATEGY property?
- COMPATIBILITY - default - do the same as it is now - inherit the environment
of the current process or use provided env map
- OVERRIDE - subprocess will inherit the environment of the current process
always and override with provided env map
Another question. What's the *current process*? Is it possible to make multi
tier subprocesses. For example main_process(main_env) -> subprocess1(none) ->
subprocess2(none)
- *subprocess1* should inherit *main_env* from *main_process*
- *subprocess1* will spawn another *subprocess2*
- the question is -> What process will *subprocess2* inherit env from?
*main_process* or *subprocess1*? It will be empty map or *main_env*?
> ABORT execvpe() crash when binaries from launcher_dir cannot be found
> ---------------------------------------------------------------------
>
> Key: MESOS-6909
> URL: https://issues.apache.org/jira/browse/MESOS-6909
> Project: Mesos
> Issue Type: Bug
> Components: agent
> Affects Versions: 1.1.0
> Reporter: Aaron Wood
> Assignee: Kevin Klues
>
> When running the Mesos agent either without --launcher_dir or with a
> --launcher_dir not pointing to the right place tasks are launched you'll get
> a crash:
> {code}
> E0111 10:50:56.665149 20924 slave.cpp:4423] Container
> '6cdd0c9b-cb29-42b0-b6cf-51f410df0f31' for executor
> '99D50FCB-ADB0-6B2A-3FC3-8A47FF178C10' of framework
> d3bc8031-29b6-4c2f-9fe3-a73c1b8b6360-0007 failed to start: Collect failed:
> Failed to setup hostname and network files: ABORT:
> (../../../3rdparty/libprocess/include/process/posix/subprocess.hpp:214):
> Failed to os::execvpe on path '/usr/local/libexec/mesos/mesos-containerizer':
> No such file or directory
> Aborted at 1484149856 (unix time) try "date -d @1484149856" if you are using
> GNU date ***
> PC: @ 0x7fc3bd418428 (unknown)
> SIGABRT (@0x51d8) received by PID 20952 (TID 0x7fc3b6007700) from PID 20952;
> stack trace: ***
> @ 0x7fc3bd7bd390 (unknown)
> @ 0x7fc3bd418428 (unknown)
> @ 0x7fc3bd41a02a (unknown)
> @ 0x47fafc _Abort()
> @ 0x47fb2a _Abort()
> @ 0x7fc3c385f092 process::internal::childMain()
> @ 0x7fc3c3864227
> _ZNSt5_BindIFPFiRKNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEEPPcS9_RKN7process10Subprocess2IO20InputFileDescriptorsERKNSC_21OutputFileDescriptorsESI_bPiRKSt6vectorINSB_9ChildHookESaISL_EEES5_S9_S9_SD_SG_SG_bSJ_SN_EE6__callIiJEJLm0ELm1ELm2ELm3ELm4ELm5ELm6ELm7ELm8EEEET_OSt5tupleIJDpT0_EESt12_Index_tupleIJXspT1_EEE
> @ 0x7fc3c38635d3 std::_Bind<>::operator()<>()
> @ 0x7fc3c3862682 std::_Function_handler<>::_M_invoke()
> @ 0x48a4b8 std::function<>::operator()()
> @ 0x7fc3c247de67 process::defaultClone()
> @ 0x7fc3c3861c40 std::_Function_handler<>::_M_invoke()
> @ 0x7fc3c3861411 std::function<>::operator()()
> @ 0x7fc3c385f8f5 process::internal::cloneChild()
> @ 0x7fc3c385d50e process::subprocess()
> @ 0x7fc3c30d318f
> mesos::internal::slave::NetworkCniIsolatorProcess::__isolate()
> @ 0x7fc3c30cf909
> mesos::internal::slave::NetworkCniIsolatorProcess::isolate()
> @ 0x7fc3c2d4db56
> _ZZN7process8dispatchI7NothingN5mesos8internal5slave20MesosIsolatorProcessERKNS2_11ContainerIDEiS6_iEENS_6FutureIT_EERKNS_3PIDIT0_EEMSD_FSB_T1_T2_ET3_T4_ENKUlPNS_11ProcessBaseEE_clESO_
> @ 0x7fc3c2d50eb8
> _ZNSt17_Function_handlerIFvPN7process11ProcessBaseEEZNS0_8dispatchI7NothingN5mesos8internal5slave20MesosIsolatorProcessERKNS6_11ContainerIDEiSA_iEENS0_6FutureIT_EERKNS0_3PIDIT0_EEMSH_FSF_T1_T2_ET3_T4_EUlS2_E_E9_M_invokeERKSt9_Any_dataOS2_
> @ 0x7fc3c380a1dd std::function<>::operator()()
> @ 0x7fc3c37eb094 process::ProcessBase::visit()
> @ 0x7fc3c37f3b26 process::DispatchEvent::visit()
> @ 0x7fc3c2244a08 process::ProcessBase::serve()
> @ 0x7fc3c37e6f50 process::ProcessManager::resume()
> @ 0x7fc3c37e3a78
> _ZZN7process14ProcessManager12init_threadsEvENKUt_clEv
> @ 0x7fc3c37f3148
> _ZNSt12_Bind_simpleIFZN7process14ProcessManager12init_threadsEvEUt_vEE9_M_invokeIJEEEvSt12_Index_tupleIJXspT_EEE
> @ 0x7fc3c37f309e
> _ZNSt12_Bind_simpleIFZN7process14ProcessManager12init_threadsEvEUt_vEEclEv
> @ 0x7fc3c37f302e
> _ZNSt6thread5_ImplISt12_Bind_simpleIFZN7process14ProcessManager12init_threadsEvEUt_vEEE6_M_runEv
> @ 0x7fc3bdc97c80 (unknown)
> @ 0x7fc3bd7b36ba start_thread
> @ 0x7fc3bd4e982d (unknown)
> {code}
> Note that this does not crash hard so the agent stays running.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
