perfSONAR User Q&A and Other Discussion

[Philip Papadopoulos <>]

Thanks, Andrew.

Patch seems to be working.. :-)

Best,

Phil

On Mon, Jul 20, 2015 at 9:22 AM, Andrew Lake <> wrote:

Hi Phillip,

I think you may have stumbled upon a bug in that graphs page. If you have 15 or less tests with data in the last month, then it shows a version of the page with a summary for throughput and/or latency. If you have more than that it just lists the endpoints. In your case, you have 14 tests with data in the last month it would appear so it wants to display a summary. Unfortunately when it goes to get the summary results, it doesn’t properly filter, and tries to grab throughput/latency/loss data for around ~230 tests that have no data in the last month in addition to the 14 that do which is causing a timeout. 

If you’d like, attached is a patch for you to try to see if it fixes things. Download the file and run the following to apply it (replace path-to-file/graphs.patch with the location where you downloaded the patch:

cd /opt/perfsonar_ps/serviceTest

patch -p1 < path-to-file/graphs.patch

No restarts or anything after that, just try reloading the page. 

Hope that helps,

Andy

On Fri, Jul 17, 2015 at 5:16 PM, Philip Papadopoulos <> wrote:

DNS looks ok.

Forward and reverse come back immediately..

But along these lines... the machine is dual homed with a public and private network.

Any perfSONAR issues that might manifest themselves because of that?

On Fri, Jul 17, 2015 at 10:56 AM, Andrew Lake <> wrote:

Hi,

It could be a number of things. The first thing to check is that your measurement archive is running and responding. As far as I can tell it is since when i go to http://ps-10g-prism.calit2.optiputer.net/esmond/perfsonar/archive/?format=json&time-range=86400 I get a quick response. 

The next thing to check will probably be your DNS configuration. After querying the archive it will do a number of DNS lookups and reverse lookups to display the host names and addresses. If your DNS configuration has a high timeout set and is timing out on one or more of the lookups things could hang. I would suspect it’s something like this as we have seen this type of thing before. You can try doing some lookups on the command-line (e.g. use the ‘host <addres>’ command or similar) of various IPs and hostnames to which you are testing and see if any of them hang. 

Hope that helps,

Andy

On Fri, Jul 17, 2015 at 12:55 PM, Philip Papadopoulos <> wrote:

See the following in /var/log/httpd/error_log

[Fri Jul 17 09:50:25 2015] [error] [client 137.110.119.117] Script timed out before returning headers: graphData.cgi, referer: http://ps-10g-prism.calit2.optiputer.net/serviceTest/psGraph.cgi

Sometimes it works, sometimes it doesn't ... what's the performance bottleneck.

Where should I look to root cause the issue?

--

Philip Papadopoulos, PhD
University of California, San Diego
858-822-3628 (Ofc)
619-331-2990 (Fax)

--

Philip Papadopoulos, PhD
University of California, San Diego
858-822-3628 (Ofc)
619-331-2990 (Fax)

--

Philip Papadopoulos, PhD
University of California, San Diego
858-822-3628 (Ofc)
619-331-2990 (Fax)