[thredds] Problems with large aggregations on JPL THREDDS?

  • To: John Maurer <jmaurer@xxxxxxxxxx>
  • Subject: [thredds] Problems with large aggregations on JPL THREDDS?
  • From: "Signell, Richard" <rsignell@xxxxxxxx>
  • Date: Sun, 17 Feb 2013 07:10:16 -0500
John,

Yes, unfortunately I think the community has frequently been seeing
these types of problems with large TDS aggregations.
The first unlucky user (or client) to access data after the metadata
in the AggregationCache has been cleared is going to have a bad
experience, waiting a long time (and perhaps timing out) before the
metadata cache is rebuilt.

 I just tried accessing the OSCAR surface current aggregation from the
JPL PO.DAAC TDS at:
<http://thredds.jpl.nasa.gov/thredds/podaac_catalogs/OSCAR_L4_OC_third-deg_catalog.html?dataset=ALL_OSCAR_L4_OC_third-deg>
and here's what happened:

- The first time (5:30 EST) I selected the opendap access link, I
waited for 5 minutes, and then I got a 503 error (service temporarily
unavailable or capacity problem).

- The second time (5:35 EST) I selected the opendap access link,
after 4.5 minutes, the OPeNDAP Access Form (finally) appeared.

- The third time (5:36 EST) I selected the opendap access link, it
responded in less than a second.

- After that, of course, the WMS also responded very quickly (and
reminded me that this is a cool dataset, godiva2 image attached)

I think the most likely problem is that the aggregationCache is
getting scoured.   On the page describing the AggregationCache setting
in threddsConfig.xml,
<https://www.unidata.ucar.edu/projects/THREDDS/tech/reference/ThreddsConfigXMLFile.html#AggregationCache>
it says:

"If you have joinExisting Aggregations, coordinate information will be
written to a cache directory....Every scour amount of time, any item
that hasnt been changed since maxAge time will be deleted. Set scour
to -1 to not scour if you have aggregations that never change."

The OSCAR dataset is not changing, so JPL could modify their
threddsConfig.xml to be:

...
<AggregationCache>
...
 <scour>-1 hours</scour>
...
</AggregationCache>

I'm cc'ing the THREDDS list as it seems a common issue.

-Rich

On Thu, Feb 14, 2013 at 4:21 PM, John Maurer <jmaurer@xxxxxxxxxx> wrote:
> Hi Rich,
> I was wondering if you've had any experience using
> JPL/PO.DAAC's THREDDS Data Server??
>
> http://thredds.jpl.nasa.gov/thredds/catalog.html
>
> There are a few things on here that my team is interested in integrating in
> my viewer (e.g. G1SST, OSTIA, OSCAR), but everything I try runs painfully
> slow or never responds at all (particularly WMS is what I've been playing
> with). My guess is that the aggregations are so large it grinds TDS nearly
> to a halt. I've tried e-mailing JPL to do some performance tuning, but they
> seem to be in denial about the problem and stopped responding. Have you had
> any success pulling in data from them??
> Aloha,
> John Maurer



-- 
Dr. Richard P. Signell   (508) 457-2229
USGS, 384 Woods Hole Rd.
Woods Hole, MA 02543-1598

Attachment: 2-17-2013 6-05-05 AM.png
Description: PNG image

  • 2013 messages navigation, sorted by:
    1. Thread
    2. Subject
    3. Author
    4. Date
    5. ↑ Table Of Contents
  • Search the thredds archives: