WESTERN REGION TECHNICAL ATTACHMENT
NO. 97-23
JULY 8, 1997



PROCEDURES FOR ENSURING RELIABLE DATA FLOW


Keith W. Meier and Jonathan Van Ausdall - NWSO Billings, MT

Overview

Forecast offices within the Western Region have been receiving gridded model data, digital satellite imagery, and other forms of model/observed data via the Western Region's Wide Area Network (WAN) for nearly two years. The data are being distributed using the Local Data Manager (LDM) software package developed by Unidata several years ago.

Although this data feed is very dependable, there are occasions when problems at a local site or the regional server contribute to missing data. Rather than accept missing data, a series of procedures was developed to allow the forecaster to proactively alleviate problems created by missing data. These procedures were developed so that the forecaster would not have to remember how to fix particular problems or possess substantial knowledge of Unix, but simply consult a flow chart and run the indicated choice from a pop-up menu by answering a few simple questions.

The series of procedures detailed below do not require the forecaster to know separate passwords or manually edit files, but simply run particular scripts to complete the desired action to fix missing data problems. In addition, these procedures also provide safeguards against local system problems that may contribute to data ingest problems.

System Problems

On occasion, local system problems may kill processes that are critical for the ingest of data (LDM) and the generation of graphics for use with Ntrans (NAG). To alleviate this, two separate safeguards have been put in place: 1) Start LDM and NAG after reboot, and 2) periodically check for LDM/NAG processes and restart if necessary.

Starting LDM and NAG after a system reboot is handled by entries in the localrc file Attachment 1 contained in the /usr/local/bin directory, as documented by Croft (1996). The entries in the localrc file replicate the interactive process of reinitializing and starting the LDM and NAG processes. Although this procedure works very well and in most instances, occasionally the LDM and NAG do not get restarted after a system reboot. An "INCLUDE" statement to call the localrc file must be included in the rc file (located in /usr/bin under HPUX 9.x and /usr/sbin under HPUX 10.x). To ensure that LDM and NAG are restarted, a script Attachment 2 is run in the cron every 30 minutes (as user Gempak) to determine if these processes are running and to restart them if they are not running. This ensures that the LDM and NAG processes do not remain inactive for a prolonged period.

Determining Necessary Action

A flowchart Figure 1 was developed to assist the forecasters in determining the appropriate action for various situations and is accessible through the office intranet. For instance, if the data may be on the local system but Ntrans graphics are not available, the forecaster would choose the pop-up menu to "Generate Graphics." Or, if the Eta model raw data were missing for the zero through 18 hour forecasts, the forecaster would first go to the WR Data Status homepage and check to see if the data was on the WR Server. If so, then the forecaster would choose the pop-up menu to "Retrieve Data and Generate Graphics". However, if the data was not on the WR Server the forecaster could also check to see if the data was on the OSO Server via the WR Data Status Homepage.

Missing Data and Data Processing Problems

With the large volume of gridded model data obtained via the WR WAN and AFOS data via the AFOS Protocol Translator (APT), it is not surprising that some data are missed or not properly processed. However, this is not necessarily a satisfying reason for the operational forecaster. In order to ensure the forecaster has some control over the data received, several procedures have been implemented to assist the forecaster in monitoring the local data feed and to actively solve data problems on the workstations.

Conclusion

The steps outlined above allow the forecaster to ensure viewable data (either through Ntrans or GARP) in situations when a data problem can be fixed locally. These local problems can be things like: data missed when a system was shutdown/rebooted; out of order graphics; LDM/NAG stopped for various reasons. The major advantages of this methodology is that the forecasters can fix many data problems very simply and all of these procedures are accomplished by users with a variety of computer skills (from little to many) by answering a few simple questions through scripts accessed from pop-up menus. This is all accomplished without having the forecasters log-in as the Gempak user (or local applications supervisor), which provides some security against inadvertently deleting important files.