Hey everyone,
So, I think my email message went out, singing the praises of how well LDM
6.13.7 worked. And, to be sure, its stability, security and much better
throughput, amongst many other things, was a really nice thing to see. And
I tested it on our AllisonHouse machines, and at COD, with no issues, on
light to moderate-duty servers.
But throw them on a server that pounds mercilessly its CPU into the ground,
and a bug was revealed in production that wasn't seen in testing: under
heavy load, a few ldmd processes hang, and won't allow the LDM to stop. You
have to killall -9 ldmd after hitting CTRL-C to kill it.
Just-released LDM 6.13.8 fixes that, tested on that mercilessly hammered
machine directly; ldmadmin stop times are anywhere from 1 to 20 seconds,
and it always stops. Furthermore, for newer AMD CPU's, LDM version 6.13.7
and older wouldn't compile. That's also addressed in 6.13.8. Finally, to
make it wroth your while to upgrade, if you type this:
ldmadmin updategempaktables
It will update your LDM (and LDM only) GEMPAK tables from Github. If you
use the NOAAport ingester, you won't get those annoying missing GRIB table
errors, as they are all updated, and future updates will not have to wait
for a new LDM version anymore!
So, do get 6.13.8 ASAP..It really does help your load average, throughput,
reliability and more. 6.13.8 fixes one bug that is in 6.13.6 and before
(but I had thought it got totally swatted in 6.13.7; thanks also for Tom
Yoksas for looking into it, as well.)
One other thing, for the newbies: you can have the fastest server in the
world, and LDM 6.13.8 on it. But if you try to feed CONDUIT or NEXRAD2
through it with limited bandwidth, nothing will help you. It definitely
will greatly cut down delays on very high volume feeds, but you need the
pipe and the server to receive those feeds successfully and timely with
high-volume feeds.
Gilbert
--
----
Gilbert Sebenste
Consulting Meteorologist
AllisonHouse, LLC
