Ganglia notes the following signature before the hang note the large incoming traffic spikes that go along with the memory use. This is consistent with a web interface page that is ingesting a large amount of data to display, like the GenomeBrowser does. Right now I see this: 2308 apache 15 0 3492m 2.8g 1796 R 26 17.8 4:00.82 perl 1334 apache 18 0 4663m 3.3g 1080 D 0 20.8 6:28.44 perl Looking at a ps with environment variable display, I see each is executing /mg-rast/FIG/metagenomics.cgi? page=MetagenomeOverview&metagenome=4440977.3 so someone should have a look at that to see what it's doing. In the meantime I'm thinking I'm going to set a memory-use limit in the apache config to 4G or so. None of our production CGIs should be using that much. On Jun 26, 2008, at 11:07 AM, Mark D'Souza wrote:

I had some pages open, and opened a number of overview pages simultaneously, which may have caused the spike.

This is something the machine should be able to handle, when it comes back I will run some tests with the pages I was opening.

The CMORE class runs next week, starting the 30th, and we will be having up to 20 students browsing the site, so we should really get this fixed.

Mark

On Thu, 26 Jun 2008, Robert Olson wrote:

...

mg-rast memory use spiked again just now; anyone happen to notice what was going on? I'm guessing it's going to be hung for a bit.

Begin forwarded message:

...

From: [email protected] Date: June 26, 2008 10:53:47 AM CDT To: [email protected] Subject: ** PROBLEM alert - mg-rast/check free memory is UNKNOWN **

***** Nagios *****

Notification Type: PROBLEM

Service: check free memory Host: mg-rast Address: 192.5.200.102 State: UNKNOWN

Date/Time: Thu Jun 26 10:53:47 CDT 2008

Additional Info:

SNMP problem - No data received from host