Re: identifying server log file when runningdistributed

Question

Hi John, I can give you some TOOL code which will get the process
id, but I do have to stray outside of Framework :-). The following
code uses classes from the SystemMonitor project which was
introduced in release 2 of Forte (this code won't work on R1):
partAgent : SystemAgent;
pidInst : ConfigValueInst;
pid : TextData;
partAgent = SystemAgent(task.Part.Agent);
pidInst = ConfigValueInst(partAgent.FindInstrument('ProcessId');
pid = TextData(pidInst.GetData);
The result is that the variable pid contains the process id in string form.
This could be converted to numeric form if needed.
If what you're really after is the partition's log file name, then the following
code will do the trick (it takes into account the differences in how the log
files are named for interpreted vs. compiled partitions):
partAgent : SystemAgent;
logFileInst : ConfigValueInst;
logFileName : TextData;
-- Get our agent and try to get the log file inst
partAgent = SystemAgent(task.Part.Agent);
logFileInst = ConfigValueInst(partAgent.FindInstrument('LogFile');
-- Interpreted partition don't have their own log file, so check
if (logFileInst = NIL) then
pidInst : ConfigValueInst;
pid : TextData;
-- We must be an interpreted partition get our pid
pidInst = ConfigValueInst(partAgent.FindInstrument('ProcessId');
pid = TextData(pidInst.GetData);
-- Build log file name
logFileName = 'forte_ex_';
logFileName.Concat(pid);
else
-- Get the name of the log file from the instrument
logFileName = TextData(logFileInst.GetData);
end;
The available agents and their instruments and commands are documented
in the manual "SystemMonitor Project". I'm at home now, so I don't have
the page numbers. Some additional agents (which were added after this
manual went to press) can be found in Tech Note #10475. Also, econsole
and escript can be handy since any instrument you can see in these tools
can be accessed from TOOL code. Hope this is of some use.
Sean
At 05:24 PM 7/30/96 -0700, John L. Jamison wrote:
>
I'd like to solicit some ideas from you folks. As many of you are probably
aware, when running in distributed mode, log output for server partitions is
written out to log files on the server partition. However it is sometimes a
trick trying to identify the process which is running your individual
partitions, and
thus knowing which log file to read.
At one client, we added a 3gl call-out to obtain the process id and return it
to the client. However this is not a good option at a new client which uses
Sequent (3gl wrappering difficult in statically linked environments such as
sequent). I am also aware that Econsole allows you to browse active
partitions and display log files, but you still have to know which active
partitions to watch.
I have not yet seen a way to programmatically obtain the process ID for a
partition within TOOL and using FrameWork classes.
What kinds of strategies are folks employing out there?
Thanks in advance,
-John
John Jamison
Sage Solutions, Inc.
353 Sacramento Street, Suite 1360
San Francisco, CA 94111
415 392 7243 x 508
[email protected]

Answer 1

Hi John,
I think that Sean Fits answered your question about TOOL code to get the PID
number. I just want to complement on the loging strategy.
There is one log file for every active partition of an application. I think
it is useful in some cases that a distributed application gets a centralized
log file to trace the exact sequential flow of processing among all the
partitions. This is useful during the initial debuging and tuning. In fact
something similar to the UNIX syslog file.
For doing so it is easy to implement a custom central log Mgr in one
partition and to have all partitions use it when needed (it doesn't prevent
to continue using the standard LogMgr in addition). This central LogMgr
automatically adds the date&time plus the node name, partition name, ... to
the log messages it receives.
The flags which apply are those of the partition where the central Log Mgr is.
Because of potential concurency requests from the several partitions
accessing the central Log Mgr, it is not possible to support the "Put" and
PutHex" methods. Only complete lines can be logged (Putline and PutHexLine).
Attached is the TOOL code of my TraceService plan that implements it.
Remark : the "Phr" in the names relate to the name of the application we
have here under development.
To use the central Log Mgr, a partition must create an object of class
PartitionLog, and then log messages must be sent to it the way you send them
to the standard LogMgr; it will manage to send them to the central Log Mgr.
At 17:24 30/07/96 -0700, John Jamison wrote:
>
I'd like to solicit some ideas from you folks. As many of you are probably
aware, when running in distributed mode, log output for server partitions is
written out to log files on the server partition. However it is sometimes a
trick trying to identify the process which is running your individual
partitions, and
thus knowing which log file to read.
At one client, we added a 3gl call-out to obtain the process id and return it
to the client. However this is not a good option at a new client which uses
Sequent (3gl wrappering difficult in statically linked environments such as
sequent). I am also aware that Econsole allows you to browse active
partitions and display log files, but you still have to know which active
partitions to watch.
I have not yet seen a way to programmatically obtain the process ID for a
partition within TOOL and using FrameWork classes.
What kinds of strategies are folks employing out there?