I compare current branch https://github.com/apache/mesos/blob/master/src/slave/slave.cpp#L3356-L3364 with 0.22 https://github.com/apache/mesos/blob/0.22.2/src/slave/slave.cpp#L2842-L2847 I think if Slave know Framework pid, why I not send ExecutorToFrameworkMessage from Slave to Scheduler directly.
FrameworkToExecutorMessage have to bypass master is because scheduler don't have Slave pid and maintain connections between every Slaves. And for >I guess that protobuf message are transmitted with TCP, so does this comment mean I have to guarantee reliability by myself even with TCP? In some cases in https://github.com/apache/mesos/blob/master/src/master/master.cpp#L3949, framework maybe could not receive ExecutorToFrameworkMessage. >If no, do we have to guarantee reliability all by ourselves Maybe could use some external systems, kafka or db? On Tue, Jan 5, 2016 at 10:47 PM, sujz.buaa <[email protected]> wrote: > Hi, all: > > > I am using mesos-0.22.0, I noticed that FrameworkToExecutorMessage is sent > along path: > Scheduler->Master->Slave->Executor, > while ExecutorToFrameworkMessage is sent along path: > Executor->Slave->Scheduler, > > > So is there some reason or benefit for bypassing master while transmitting > ExecutorToFrameworkMessage? > > One more question, FrameworkToExecutorMessage and ExecutorToFrameworkMessage > are instantiated > in function SendFrameworkMessage, > declaration of SendFrameworkMessage in include/mesos/scheduler.hpp and > include/mesos/executor.hpp: > // Sends a message from the framework to one of its executors. These > // messages are *best effort*; do not expect a framework message to be > // retransmitted in any reliable fashion. > virtual Status sendFrameworkMessage( > const ExecutorID& executorId, > const SlaveID& slaveId, > const std::string& data) = 0; > > > I guess that protobuf message are transmitted with TCP, so does this comment > mean I have to guarantee reliability by myself even with TCP? What's special > for these > > two messages compared with other protobuf messages, If no, do we have to > guarantee reliability all by ourselves? > > Thank you very much and best regards ! > -- Best Regards, Haosdent Huang
