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

20020925: LDM 5.2.1 source distribution available (cont.)

>From: "Arthur A. Person" <address@hidden>
>Organization: PSU
>Keywords: 200207291959.g6TJxE922257 LDM 5.2 pqsurf large queue

Hi Art,

>I've upgraded our ldm and started operation.  We now run RedHat 7.3 fully
>patched, large files (3G queue) and ldm 5.2.1.  Things seem to be running
>okay at the moment.

Excellent.

>I do notice one potential problem, though...  we seem
>to be accumulating:
>
>tcp        0      0 ldm.meteo.psu.edu:33803 flip.unidata.ucar.e:ldm TIME_WAIT
>
>when I do "netstat -a" which I assume is from the rtstat program.  Is this
>a problem, or will it be self limiting?  Currently, there are 46 lines
>like these.  I'll let you know if I notice anything else.

This should be self limiting, but I will check with Chiz to make sure
that there is nothing amiss on rtstats.unidata.ucar.edu.

>What kind of statistics are you accumulating with rtstats and what are the
>plans for their use?  Is there a way that I can use rtstats to view these
>statistics myself (...I haven't read the man page yet, perhaps the
>answer's in there?).

The "realtime statistics" (rtstats) will eventually replace the
reporting from pqbinstats when all of the kinks are worked out and all
sites have upgraded to LDM 5.2+.  Chiz setup a system that automatically
creates graphs of the rtstats reportings for each site.  To check out
the results, visit:

http://www.unidata.ucar.edu/staff/chiz/latency/stats

This page has two links:

Statistics by Feedtype

Statistics by Host

Either link will get you to output from your and other IDD nodes. I
always use 'Statistics by Host' since I am typically trying to get a
feeling for how well a particular IDD node is functioning.

When you click on the 'Statistics by Host' link, you will see a page
that lists all of the IDD nodes reporting rtstats.  Click on
ldm.meteo.psu.edu, and you will see links for each feedtype that it is
receiving.  After your reporting has been up for a while, the graphs
will settle down and begin reflecting about a week's worth of latency
values for the particular feed.

We have found the rtstats output to be extremely valuable in
troubleshooting feed problems, and for generally getting an idea of how
data is flowing through the IDD.  The plan, of course, is to get
everyone participating in the IDD to upgrade to LDM 5.2+ and run
rtstats.

Latest information:

Chiz found and fixed a memory leak in rtstats late on Monday.  The code
modifications will be included in the next LDM release.  Stay tuned
for an announcement of its availability.

Please let us know what you think about the rtstats graphs.

Tom