104 lines
4.0 KiB
Plaintext
104 lines
4.0 KiB
Plaintext
|
|
pestat - a tool to monitor resources on all PBS nodes
|
|
-----------------------------------------------------
|
|
|
|
The pestat code contacts every node served by the given
|
|
PBS server and retrieves resource information such as
|
|
CPU load and memory usage (any resource information can be
|
|
programmed into the code, if you need it).
|
|
A single summary line is printed for each node.
|
|
|
|
The latest version of this code is available from
|
|
ftp://ftp.fysik.dtu.dk/pub/PBS/
|
|
|
|
The code has been tested on the following architectures:
|
|
o Linux (the default)
|
|
o Compaq Tru64 UNIX 4.0F
|
|
|
|
Example output (from Linux):
|
|
|
|
# pestat
|
|
node state load pmem ncpu mem resi usrs jobs jobids
|
|
p01 free 0.15 511 1 0 0 6/3 0
|
|
p02 free 0.00 511 1 0 0 0/0 0
|
|
p03 free 0.00 511 1 0 0 0/0 0
|
|
p04 free 0.00 511 1 0 0 0/0 0
|
|
p05 free 0.00 511 1 0 0 0/0 0
|
|
p06 free 0.00 511 1 0 0 0/0 0
|
|
p07 excl 0.73 511 1 0 0 1/1 1 219
|
|
p08 excl 0.94 511 1 0 0 1/1 1 217
|
|
p09 excl 0.99 511 1 0 0 1/1 1 217
|
|
p10 excl 0.99 511 1 0 0 1/1 1 217
|
|
|
|
Example output (from Tru64 UNIX):
|
|
|
|
# pestat
|
|
node state load pmem ncpu frmem ubcmem usrs jobs jobids
|
|
asrv free 1.24* 640 1 274 62 5/4 0
|
|
acmp free 0.00 1536 1 1139 75 9/9 0
|
|
a01 excl 0.00* 512 1 311 24 1/1 1 14371
|
|
a02 excl 0.01* 512 1 314 24 1/1 1 14371
|
|
a03 excl 1.00 512 1 323 24 1/1 1 14263
|
|
a04 excl 0.97 512 1 97 24 1/1 1 14298
|
|
a05 excl 1.00 512 1 309 24 1/1 1 14371
|
|
a06 excl 0.60 512 1 361 17 1/1 1 14582
|
|
a07 excl 0.56 512 1 355 18 1/1 1 14582
|
|
a08 excl 1.00 512 1 143 24 1/1 1 14598
|
|
a09 free 0.00 512 1 438 14 0/0 0
|
|
|
|
Some nodes have an asterisk (*) next to the "load" column.
|
|
This indicates a node whose load is "unexpected", i.e., a
|
|
free node with a high load or a busy node with low load.
|
|
The "acceptable" load-range for a busy node has arbitrarily
|
|
been taken as 0.5-1.5, but this can of course be changed in
|
|
the code (look for the loadave variable).
|
|
|
|
Installation
|
|
------------
|
|
|
|
Edit the Makefile: Change PBSHOME to point to your PBS
|
|
source directory, since pestat.c needs several header files
|
|
from the PBS source distribution.
|
|
|
|
Look at the supported architectures and uncomment the relevant lines.
|
|
Check that the include and lib paths point to the correct
|
|
directory.
|
|
|
|
Type "make" to generate the executable "pestat". Run pestat
|
|
to verify the correct operation. Copy pestat to a directory
|
|
in your PATH, such as /usr/local/bin.
|
|
|
|
Linux notes:
|
|
------------
|
|
|
|
The OpenPBS 2.3 RPM-package installs include and lib files
|
|
in the directory /usr/pbs. You will also need to unpack
|
|
the source distribution, since pestat.c needs several header files
|
|
from the PBS distribution.
|
|
|
|
Compaq Tru64 UNIX notes:
|
|
------------------------
|
|
|
|
The Makefile assumes that you have copied the relevant PBS files
|
|
to /usr/local/include and /usr/local/lib, so edit this if
|
|
necessary. One snag is that you can't link with "-lnet"
|
|
because that would incorrectly pick up /usr/shlib/libnet.so.
|
|
|
|
The pestat.c code has some "#ifdef TRU64" sections that assume
|
|
that the resources named "freemem" and "ubcmem" are available.
|
|
The latter refers to the Tru64 UBC (Unified Buffer Cache)
|
|
memory reservation. These resources requires that you have
|
|
compiled PBS with the MOM-patches written by Mohan <mkannapa@ford.com>.
|
|
These patches should become available on http://www.mcs.anl.gov/openpbs/.
|
|
|
|
Author:
|
|
-------
|
|
|
|
Originally written by David.Singleton@anu.edu.au
|
|
(ANU Supercomputer Facility, Australian National University).
|
|
|
|
Questions: Ole Holm Nielsen, Ole.H.Nielsen@fysik.dtu.dk
|
|
(Technical University of Denmark).
|
|
|
|
Please send bug reports, support for other architectures, etc. to Ole.
|