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

Re: Data reset



Harry Edmon wrote:
> 
> There were a number of resets last night.  Can you check the logs on thelma?
> Also, the log entries you sent me seem to truncate at 80 characters (I don't 
> see
> the full header).  My working assumption, which is the one I had when we bad a
> problem between sunny and air (which is not occuring currently) is that the
> product being sent at the time of the timeout is never resent.
> 

Wow, on Apr 19 I see exits from sunny at:

02:08:02
02:58:59
03:18:04
03:36:53
03:41:33
04:04:02
04:23:34
04:41:53
05:02:18
07:03:01
07:53:06
09:28:02
10:28:01
12:38:07
15:13:23
18:13:01
18:38:42
19:18:37
20:18:51

Yes, the log entries are truncated in the log, which is bad for the
CONDUIT products.  (Maybe a useful change would be to log the rightmost
portion of such a long product ID.)  

I was assuming differently, that the products that were dropped were not
the products on which the timeout occured.  I assumed this because I
have not heard other reports of dropped products on disconnects and thus
I assumed that the logic to resend a product that was not completely
sent was good.   I'm thinking that the problem lies in the ordering of
the products in the queue.  With the CONDUIT feed being broken up into
multiple connections, I'm wondering if that increases the liklihood of
products being ordered nonchronologically in the queue, resulting in
more dropped products.  This is why I asked if you had noticed dropped
products on other feeds besides CONDUIT.  Have you?

Anne
-- 
***************************************************
Anne Wilson                     UCAR Unidata Program            
address@hidden                 P.O. Box 3000
                                  Boulder, CO  80307
----------------------------------------------------
Unidata WWW server       http://www.unidata.ucar.edu/
****************************************************