Re: [thredds] aggregation caching: cache cleared when Tomcat restarted?

• To: thredds@xxxxxxxxxxxxxxxx
• Subject: Re: [thredds] aggregation caching: cache cleared when Tomcat restarted?
• From: "John Maurer, IV" <jmaurer@xxxxxxxxxx>
• Date: Thu, 06 Jan 2011 08:07:22 -1000

Hi John,

> TDS has to rebuild its internal structures upon restart, though im 
> surprised its particularly slow since the agg coordinate values 
> are cached.
>
> you previously indicated ncml like:
> 
> <aggregation dimName="time" type="joinExisting" 
> timeUnitsChange="true"><scan 
> location="/export/lawelawe1/model/tide/netcdf_data/mhi/elev/" 
> suffix=".nc"/>
> </aggregation>
> 
> since theres no "recheckEvery" attribute, once scanned, it should not 
> update except on reboot. Its likely the bug is in that logic. 
> Ill try to recreate it, or add some diagnostics for you to run.
> 
> can you first double check that the server is not being 
> restarted? look 
> in threddsServlet.logs, the "reqSeq" field will start over on restart.

I have double checked and the server has not been restarted since January 3rd 
at 3pm. The server does not restart without my manually controlling it, so 
there shouldn't be any restarts that I'm unaware of. Thank you for looking into 
a possible bug. Would it help to add a "recheckEvery" attribute? Another piece 
of the puzzle I discovered this morning while poking through my logs:

2011-01-03T15:03:38.619 -1000 [       934][       1] ERROR -  
thredds.servlet.ThreddsConfig - ThreddsConfig: param  AggregationCache.scour 
not udunit time Can't convert from unit "-1.0" to  unit "s"

Apparently it does not like <scour>-1</scour>, despite what I read in the 
documentation. Could that be the problem? Maybe only part of the problem, since 
I've removed AggregationCache entirely in previous tests and still get 
recurring slowness. Lemme know...
Thanks!,
John Maurer