Service is not stable!

[Klortho]

Try going here, http://eutils.org/data/pubmed?format=rdf, and hitting F5 a few times. Invariably, you will get 500 Internal Server Error

[Klortho]

Testing with accessing the http://eutils.org/data/pubmed?format=rdf.

Using wireshark to captured traffic that shows the error. This dumps all packets between MarkLogic and eutils.ncbi.nlm.nih.gov:

sudo tshark -i eth0 -f 'host 130.14.29.110'

I then created a mockup cgi, that delivers the same XML, at http://chrisbaloney.com/einfo-pubmed.xml, and captured traffic with

sudo tshark -i eth0 -f 'host 69.255.231.180'

[Klortho]

See these network traffic traces: https://github.com/Klortho/eutils-org/tree/master/issues/13.

This problem manifests when you hit a page like http://eutils.org/data/pubmed?format=rdf twice in quick succession. I can't figure out what the source of the error is.

When you hit that page, MarkLogic does a request to http://eutils.ncbi.nlm.nih.gov/entrez/eutils/einfo.fcgi?tool=eutils.org&email=voldrani@gmail.com&db=pubmed (see the source).

Something weird is going on between MarkLogic and eutils. It never fails when I substitute a dummy cgi for eutils, at http://chrisbaloney.com/einfo-mock.cgi.

I made three network packet traces with wireshark:

• eutils.good.packets.txt - shows what happens when you wait long enough after the first request, so that both requests succeed.
• eutils.bad.packets.txt - shows what happens when you hit the URL twice in succession; the second request fails.
• chrisbaloney.packets.txt - shows what happens with the mockup cgi, both when you wait after one request, and when you don't.

[Klortho]

Also, note that the bad requests do not show up in applog. http://mini.ncbi.nlm.nih.gov/1k6ae shows three requests, all of which completed successfully. The fourth one failed, and doesn't show up.