Determining if process is alive on a different node

Serge Aleynikov serge@REDACTED
Mon Aug 15 21:51:55 CEST 2005


Sorry, after rereading your message I realized I responded too quickly.  :-(

Yes the
rpc:call(node(Pid), erlang, is_process_alive, [Pid]) would be the 
logical way to do it, but is it reliable?

What if the node(Pid) node got bounced since the time we captured the 
Pid.  The Pid value might've been reassigned to another process, as in 
the example below:

(a@REDACTED)18> register(test, self()). 


(b@REDACTED)1> net_adm:ping(a@REDACTED).

(b@REDACTED)2> {test, a@REDACTED} ! self().

(a@REDACTED)2> receive M -> M end.

BREAK: (a)bort (c)ontinue (p)roc info (i)nfo (l)oaded
        (v)ersion (k)ill (D)b-tables (d)istribution
~/tmp>erl -sname b
Erlang (BEAM) emulator version 5.4.8 [source] [hipe] [threads:0]
Eshell V5.4.8  (abort with ^G)
(b@REDACTED)1> net_adm:ping(a@REDACTED).

(a@REDACTED)5> rpc:call(b@REDACTED, erlang, is_process_alive, 

This last statement is not true, since it is not the same process on 
node b that originated the first message.  The Pid that was originally 
recorded was reassigned to another process after the node was bounced.

I believe it is more reliable to use erlang:monitor/2 calls at a time 
the Pid is captured, and watch for {'DOWN', ...} messages.



Joel Reymont wrote:
> Serge's use of use of node(Pid) made me thing that the following  would 
> work:
> rcp:call(node(pid), erlang, is_process_alive, [Pid]).
> Any drawbacks?
>     Thanks, Joel
> On Aug 15, 2005, at 9:06 PM, Serge Aleynikov wrote:
>> node(Pid) returns the node that owns the Pid.
>> Node = node(),
>> case node(Pid) of
>> Node -> local;
>> _    -> remote
>> end.

Serge Aleynikov
R&D Telecom, IDT Corp.
Tel: (973) 438-3436
Fax: (973) 438-1464

More information about the erlang-questions mailing list