|
David,
I can not really tell from your message the exact problem
you might be experiencing. I would be curious to find out what
version of the V2.1 Schduler server you are using because the
agent you are referencing communicates via TCP/IP. Later versions
of Server communicate to agents via TCP/IP. I think I am stating
this correctly. This might account for the problem you are experiencing.
There is a later version of the agent V2.1B-04.
The following are some suggestions for Agent Tuning:
In order to accommodate the diversity of environmental conditions that
may exist at different installation sites, this version includes the
ability to tune the communication code that enables the Scheduler and
agent to communicate.
Sites with slow networks, unreliable networks, heavily burdened nodes,
or heavy spikes in CPU utilization may need to adjust these parameters.
The tuning parameters use the same names for OpenVMS and for the agent.
On OpenVMS, these parameters may be logical names or symbols. If they
are system logical names, they will affect the NSCHED, SCHED$LISTENER,
RETRY, and user interface processes. If it is desired to have different
parameters for these processes, use symbols or process logical names
created in the process's startup file. On UNIX systems the parameters
are environment variables; use the setenv command and make sure that
all environment variable names are uppercase.
The following table lists parameters along with their default,
minimum, and maximum settings:
____________________________________________________________________________
Parameter_________________Default___Minimum___Maximum_______________________
SCHED_PACKET_WAIT 2 (sec.) 1 (sec.) 60 (sec.)
SCHED_MAX_RETRIES 3 0 32
SCHED_PENDING_DELAY 60 1 (sec.) -
__________________________(sec.)____________________________________________
The debug level can be set via startup parameters for the agent.
I suggest you reference the server release for 2.1B under section
3 there is a full description. However, the agent accepts the
following parameters at startup time:
argv 1 = port number (default = 5841)
argv 2 = default directory (default value is passed 0)
argv 3 = debug level --> use 5
argv 4 = detached --> use 1
In all truthfulness, I would suggest the latest version of the
agent because the TCP/IP is much better and needless to say the
Server latest version is also.
I know this doesn't directly answer the problem. But I hope it helps.
|
| Sorry,
The base note was a bit terse. I am actually running the Digital Unix
Scheduler server. Info below.
>>> digital unix box
Agent Scheduler Version Agent Start Time
----- ----------------- ----------------
dec002 V3.0-02 25-Mar-96 15:57:19
Current Job Count : 0 Max Jobs : 64
CPU Weight : 100 I/O Weight : 0
Memory Weight : 0 LBAL Interval : 15
Load Rating : 24031
Debug : FALSE Default Run Priority : 0
SCHEDULE> sho agent rs01 display full
The agent only dies on some HP-UX boxes. I will turn on the debug stuff and
post results.
Thanks,
David
|