Hi All,
Just want to update on how test runs has been so far this week.
The workflows are running good for most of the configurations, we
certainly came down quite a bit from 100% success rate but now sure
exactly how much, hopefully not below 90%. With good amount of testing
from all of you we have a found a problem which prevented you from
launching a workflow this morning. Restarts of couple of services helped
and everything is fine now but we still did not resolve the problem and
will continue to look why the unscheduled and unwanted down time happened.
If you have noticed yesterday between morning and 3pm (EST) workflows
failed due to some change in a software. We reverted that change and all
failed workflows were resurrected late in the evening and finished
successfully but not sure if they did a forecast or a post cast. The
consequence of this will be duplicates of data in your workspace for the
experiments created yesterday.
We have seen some WRF runs going unstable in past 2 days, not sure if
they gave workflow success or failed status. I did not completely
understand the reasons Dan Weber and Brian Jewett have analyzed but you
guys might, which is "Looking at the plots Brian generated from the
earlier run, convection starts on the southern boundary, which usually
means doom for a forecast, especially if it is on an inflow condition
(which is the case for the run). No convection was visible in the the
ADAS analysis during the forecast area. It seems that the main problem
is the placement of the lateral boundary in a region that is mountainous
and where convection developed. This is the one of the problems with
limited domain forecasts, such as the 5km 201x201x51 grid" -- We still
need to parse this to english and then make the changes to data search
or namelists.
All of this aside, please continue to use our resource reservations at
NCSA from 1pm to 6pm (EST) and test and report problems as usual.
Happy Testing,
Suresh
==============================================================================
To unsubscribe leadusers, visit:
http://www.unidata.ucar.edu/mailing-list-delete-form.html
==============================================================================
