headers
Ray Plante | 29 Jan 2008 19:08
Picon

detection results not persisting

Serge, K-T, and Tim,

Please have a look at the results in /lsst/DC2root/rlp0094.  The logs 
messages seen on the web (e.g.
http://dev.lsstcorp.org/cgi-bin/showSlice.cgi?rlp0094,20) show a failure 
to persist the results from the detection stage.  Could this be due to the 
recent change in the schemas?  Does detection need to be updated?

Also, eventually, we see an error from MySQL complaining of too many 
connections.  Do we need to put in some "finally" code somewhere to 
clean up open connections in case of error?

thanks,
Ray
Permalink | Reply |
headers
Kian-Tat Lim | 29 Jan 2008 22:16
Picon
Favicon

Re: detection results not persisting

Ray wrote:
> Also, eventually, we see an error from MySQL complaining of too many 
> connections.  Do we need to put in some "finally" code somewhere to 
> clean up open connections in case of error?

Thanks to Serge, this should be fixed (although not in the most elegant
way) in mwi [4050] and [4051].

--

-- 
Kian-Tat Lim, LSST Data Management, ktl <at> slac.stanford.edu
Permalink | Reply |
headers
Kian-Tat Lim | 29 Jan 2008 20:57
Picon
Favicon

Re: detection results not persisting

Ray asked:
> a failure to persist the results from the detection stage.  Could this
> be due to the recent change in the schemas?

	Yes.  The template table still has ampExposureId instead of
exposureId.  Somehow you have the wrong .sql file for the table setup.

> Does detection need to be updated?

	I don't think so.  You may have the wrong version of dc2pipe
setup.

> Also, eventually, we see an error from MySQL complaining of too many
> connections.  Do we need to put in some "finally" code somewhere to
> clean up open connections in case of error?

	Jacek may have upped the number of connections, but we shouldn't
be running into this with 20 slices.  It does look like something is not
getting closed somewhere.  I will research this.

--

-- 
Kian-Tat Lim, LSST Data Management, ktl <at> slac.stanford.edu
Permalink | Reply |
headers
Ray Plante | 29 Jan 2008 22:31
Picon

Re: detection results not persisting

On Tue, 29 Jan 2008, Kian-Tat Lim wrote:
> You may have the wrong version of dc2pipe
> setup.

That is indeed what the problem was.  We appear to be running fine now 
with rlp0095.

thanks,
Ray
Permalink | Reply |
headers
Serge Monkewitz | 29 Jan 2008 20:49
Picon

Re: detection results not persisting

On Tue, 29 Jan 2008 10:08:38 -0800, Ray Plante  
<rplante <at> poplar.ncsa.uiuc.edu> wrote:

> Serge, K-T, and Tim,
>
> Please have a look at the results in /lsst/DC2root/rlp0094.  The logs
> messages seen on the web (e.g.
> http://dev.lsstcorp.org/cgi-bin/showSlice.cgi?rlp0094,20) show a failure
> to persist the results from the detection stage.  Could this be due to  
> the
> recent change in the schemas?  Does detection need to be updated?

The fw unit tests pass against the newest schema, and ampExposureId isn't  
set or used by detection (and this is the only column that changed in the  
DIASource schema). Is it possible that the latest version of fw wasn't  
declared current for the run?
Permalink | Reply |
headers
Jacek Becla | 29 Jan 2008 20:02
Picon
Favicon

Re: detection results not persisting

Ray,

I reconfigured mysql on lsst-dev10 to avoid the
"too many connections" problem. It used to be
configured for max 100 connections, now it is 1K.
I did that without taking down the server.

Jacek

Ray Plante wrote:
> Serge, K-T, and Tim,
> 
> Please have a look at the results in /lsst/DC2root/rlp0094.  The logs 
> messages seen on the web (e.g.
> http://dev.lsstcorp.org/cgi-bin/showSlice.cgi?rlp0094,20) show a failure 
> to persist the results from the detection stage.  Could this be due to the 
> recent change in the schemas?  Does detection need to be updated?
> 
> Also, eventually, we see an error from MySQL complaining of too many 
> connections.  Do we need to put in some "finally" code somewhere to 
> clean up open connections in case of error?
> 
> thanks,
> Ray
> 
> _______________________________________________
> LSST-data mailing list
> LSST-data <at> lsstcorp.org
> http://listserv.lsstcorp.org/mailman/listinfo/lsst-data
(Continue reading)

Permalink | Reply |

Gmane