Closed divergentdave closed 7 years ago
According to gdb, one process was reading from an SSL socket.
Here's what I got from /proc/<PID>/net/tcp
:
sl local_address rem_address st tx_queue rx_queue tr tm->when retrnsmt uid timeout inode
269: F2391FAC:DBD0 2015E0A5:01BB 01 00000000:00000000 00:00000000 00000000 1000 0 1629135 1 ffff8800373f6900 708 4 26 2 -1
Edit: local_address is 172.31.57.242:53467 remote_address is 165.224.21.32:47873 I may well have gotten the port number the wrong way around, but oh well. The remote host is ed.gov, so I'll look into that more.
I killed one of the hung processes, and this was the tail end of the log.
https://www2.ed.gov/about/offices/list/oig/otheroigproducts.html
[other][2016-09-26][x21q0001]
report: education/2016/x21q0001/report.pdf
text: education/2016/x21q0001/report.txt
data: education/2016/x21q0001/report.json
[other][2016-08-15][coveredsystems088152016]
report: education/2016/coveredsystems088152016/report.pdf
text: education/2016/coveredsystems088152016/report.txt
data: education/2016/coveredsystems088152016/report.json
[other][2016-02-29][scrareport02292016]
report: education/2016/scrareport02292016/report.pdf
text: education/2016/scrareport02292016/report.txt
data: education/2016/scrareport02292016/report.json
https://www2.ed.gov/about/offices/list/oig/specialreportstocongress.html
https://www2.ed.gov/about/offices/list/oig/testimon.html
https://www2.ed.gov/about/offices/list/oig/ireports.html
bash: line 1: 31525 Terminated python igs
For some reason scrapers are intermittently hanging. I noticed this yesterday when enough hung processes had piled up to exhaust memory, and this started showing up on the dashboard when subprocesses wouldn't fork anymore.
I spun up a fresh machine last night, and right now there are two lingering scraper processes. (see below) In the same time, six cron jobs ran successfully.
strace
says both processes are reading from a socket, so presumably we are doing some network access without a timeout.