[MESOS-1328] Slave should not rely on a checkpointed forkedPid

• Type: Bug
• Status: Resolved
• Priority: Minor
• Resolution: Duplicate
• Affects Version/s: 0.20.0
• Labels: checkpoint recovery slave
• Assignee: Unassigned
• Reporter: Till Toenshoff
• Created: 08/May/14 01:12
• Updated: 02/Jul/15 18:43
• Resolved: 16/May/14 23:45

Description

Within slave.cpp, line 3309 we see the following:

    CHECK_SOME(run.get().forkedPid)
      << "Failed to get forked pid for executor " << state.id
      << " of framework " << id;

For the ExternalContainerizer, that is a problem. The EC should not (actually can not) checkpoint a forked PID of an executor as that is state-keeping-responsibility of the ECP (EC-counterpart).

Issue Links

• duplicates: MESOS-923 Containerizers should persist their own recovery state.