Very Computer

> Hi all,
> I working on an application on linux which involves creating redundant
> processes for high availability. roughly it is something like this:

> clients-->Process A------> Process B--->Process C
> the requirement is that each process A,B,C should have a slave process to
> take over in case something happen to it.

> clients-->Process A------> Process B--->Process C
> the requirement is that each process A,B,C should have a slave process to
> take over in case something happen to it. so, my quesion is: how should i
> go about creating slave processes ? should they the same as the master
> processes but with a flag to indicate that they are slaves, and once they
> take over change the flag to indicate that they are the masters now and
> create new slave processes ? how about the communication between the master
> and the slave ?

> I just need pointers in the right direction.

>> clients-->Process A------> Process B--->Process C
>> the requirement is that each process A,B,C should have a slave process to
>> take over in case something happen to it. so, my quesion is: how should i
>> go about creating slave processes ? should they the same as the master
>> processes but with a flag to indicate that they are slaves, and once they
>> take over change the flag to indicate that they are the masters now and
>> create new slave processes ? how about the communication between the
>> master and the slave ?

>> I just need pointers in the right direction.

> The way I'd go about this is, for one such process A, have a
> "supervisor" process, and the "worker" process.  The supervisor process
> is the first process started.  It forks off a child, which becomes the
> worker, then waits for the child to die.  When this happens, it forks
> off another child to replace the previous one... lather, rinse, repeat.