Back Midas Rome Roody Rootana
  Midas DAQ System  Not logged in ELOG logo
Entry  07 May 2009, Konstantin Olchanski, Info, RPC.SHM gyration 
    Reply  02 Jun 2009, Konstantin Olchanski, Info, RPC.SHM gyration 
       Reply  04 Jun 2009, Stefan Ritt, Info, RPC.SHM gyration 
    Reply  20 Nov 2009, Konstantin Olchanski, Info, RPC.SHM gyration 
Message ID: 586     Entry time: 02 Jun 2009     In reply to: 570     Reply to this: 589
Author: Konstantin Olchanski 
Topic: Info 
Subject: RPC.SHM gyration 
> When using remote midas clients with mserver, you may have noticed the zero-size .RPC.SHM files 
> these clients create in the directory where you run them. These files are associated with the semaphore 
> created by the midas rpc layer (rpc_call) to synchronize rpc calls between multiple threads. This 
> semaphore is always created, even for single-threaded midas applications. Also normally midas 
> semaphore files are created in the midas experiment directory specified in exptab (same place as 
> .ODB.SHM), but for remote clients, we do not know that location until we start making rpc calls, so the 
> semaphore file is created in the current directory (and it is on a remote machine anyway, so this 
> location may not be visible locally).
> 
> There are 2 problems with these semaphores:

A 3rd problem surfaced - on SL5 Linux, the global limit is 128 or so semaphores and on at least one heavily used machine that hosts multiple 
experiments we simply run out of semaphores.

For "normal" semaphores, their number is fixed to about 5 per experiment (one for each shared memory buffer), but the number of RPC 
semaphores is not bounded by the number of experiments or even by the number of user accounts - they are created (and never deleted) for 
each experiment, for each user that connects to each experiment, for each subdirectory where the each user happened to try to start a 
program that connects to the each experiment. (to reuse the old children's rhyme).

Right now, MIDAS does not have an abstraction for "local multi-thread mutex" (i.e. pthread_mutex & co) and mostly uses global semaphores 
for this task (with interesting coding results, i.e. for multithreaded locking of ODB). Perhaps such an abstraction should be introduced?

K.O.
ELOG V3.1.4-2e1708b5