advforecast.php - forecast time confusion - NWS server farm issue?

I’ve been playing with Tom/Ken/Mike’s advforecast.php script and seeing it generate forecasts with two slightly different NWS stamps as the page is refreshed with “force” on. I thought this was due to something I had setup incorrectly but couldn’t figure out, but now I see that Ken’s site will show the same thing.

It’s 10:15 AM here in sunny California. If I persistently refresh Ken’s advforecast2.php I can get:


The format of the captured forecast.txt files (attached) are slightly different too.

Has anyone else noticed this?

forecast 9_25.TXT (4.79 KB)

forecast 9-41.TXT (4.82 KB)

It does seem to be NWS server difference. The 9:25 one has headers

HTTP/1.1 200 OK
Date: Wed, 10 Mar 2010 18:19:34 GMT
Server: Apache
Cache-Control: max-age=900
Expires: Wed, 10 Mar 2010 18:34:34 GMT
Content-Length: 4621
Connection: close
Content-Type: text/html; charset=UTF-8

while the 9:41 one has

HTTP/1.0 200 OK
Age: 0
Date: Wed, 10 Mar 2010 18:21:27 GMT
Content-Length: 4621
Content-Type: text/html; charset=UTF-8
Expires: Wed, 10 Mar 2010 18:31:27 GMT
Cache-Control: max-age=600
Server: Apache
Via: 1.1 hyacinth (NetCache NetApp/6.0.3)

Looks like there’s a transparent proxy cache server set involved which sometimes you get, sometimes you don’t.

The DNS for shows:

03/10/10 11:21:32 dns
Canonical name:

03/10/10 11:21:54 dig @
Dig [email protected] ...
Non-authoritative answer
Recursive queries supported by this server
 Query for type=255 class=1 PTR (Pointer)

so it’s not a round-robin DNS… it appears to be a transparent proxy that is sometimes engaged for queries.

Best regards,

Same thing happening here- 2:43pm now refreshing will either give you:

National Weather Service Forecast for: Friend NE
Issued by: National Weather Service WFO Omaha, NE
Updated: 2:19 pm CST Mar 10, 2010


National Weather Service Forecast for: Friend NE
Issued by: National Weather Service WFO Omaha, NE
Updated: 2:34 pm CST Mar 10, 2010


Thanks for the info and confirmation guys :smiley: In practice I doubt it has any real impact but I like to thoroughly test what I implement, and when I can’t get a consistent answer that’s not good. You would think anyone building a load balancing system like this would automatically ensure that all the data sources were in sync, but I guess not :?

Thanks also to Mark/Snoqualmie WX for his help trying to figure this one out.