[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

20050812: McIDAS, LDM, and GEMPAK installs at USF (cont.)



>From:  "Happel, Shelly" <address@hidden>
>Organization:  USF
>Keywords:  200508121849.j7CInOjo025553 Unidata package installations

Hi Shelly,

re: installing GEMPAK
>Thanks Tom! That sounds great. Yes, if it's possible for you to install
>it, I believe we want it since both our grad student and new faculty
>hire are the most familiar with that software. Thanks again,

I installed the latest release of GEMPAK after creating the user
'gempak' and HOME directory /home/gempak.  I also configured the LDM to
run GEMPAK decode processes and setup data scouring for the data that
those decoders create.  I _think_ that the data scouring is correct and
complete, but someone there needs to keep an eye on disk use and make
adjustments if/when needed (otherwise you _will_ run out of disk.

One comment.  metlab went from being idle to being busy.  You may find
that it does not have enough umph to ingest, file, decode, and use as a
workstation with the volume of data it is processing.  You should keep
an eye on load averages, disk use, etc. to see if you are asking the
machine to do too much.  To help monitor system performanc, I installed
a Tcl script that is run by cron from the LDM account.  It creates its
output in the ~ldm/logs directory in a file named usf.uptime.  Here is
what the file contents loooks like:

20050812.2215   1.48  1.36  1.01    0   3   3   3781    4M    2M
20050812.2216   1.02  1.24  0.99    0   3   3   3784    4M    2M
20050812.2223   0.71  0.54  0.73    0   3   3   4095   42M     0
20050812.2224   3.77  1.47  1.04    0   3   3   3713    5M     0
20050812.2225   3.60  1.88  1.20    0   3   3   3454    4M     0
20050812.2226   1.83  1.68  1.18    0   3   3   3445    4M     0
20050812.2227   3.02  2.04  1.33    0   3   3   3360    5M     0
20050812.2228   1.65  1.85  1.31    0   3   3   3295    5M     0
20050812.2229   0.81  1.56  1.24    0   3   3   3290    4M     0

The info pieces mean:

20050812.2215    - date.time in UTC
1.48  1.36  1.01 - 1, 5, 15 load averages
0   3   3        - # downstream feeds, # upstream requests, # total
3781    4M    2M - age of oldest product in queue, free memory, swap used

I set the LDM queue size to 500 MB in ~ldm/etc/ldmadmin-pl.conf.  This
may need to be increased if data decoding can't keep up with the volume
of data you are ingesting.

In order to make the data tree layout look like Unidata
recommendations, I make a link from /var/data/ldm to /data/ldm.  This
link must be maintained for the McIDAS-XCD decoding ADDE serving of
data to continue working.

I also setup ntp to keep the clock set.  You can see that metlab is
able to receive the data coming in so far through the real time
statistics pages for USF:

Real Time Stats HomePage
http://my.unidata.ucar.edu/content/software/idd/rtstats/index.html

Statistics by Host
http://my.unidata.ucar.edu/content/software/idd/rtstats/siteindex.php

metlab.cas.usf.edu
http://my.unidata.ucar.edu/content/software/idd/rtstats/siteindex.php?metlab.cas.usf.edu

OK, now it is up to you/USF users to check things out and make adjustments
if/where necessary.

Cheers,

Tom

>Date: Tue, 16 Aug 2005 12:56:43 -0400
>From: "Happel, Shelly" <address@hidden>

>Hi Tom,

>Thank you SO much for all of your work on the system. We'll look over it
>and I'll get back with questions. Thanks again!! We all really
>appreciate it here.

>Take care,

>s.