1/26/2015 Notes

DocDB Event


Alec, Zukai, Jan, Pengfei, Yuri, Craig G., Craig D., and Martin (taking notes)

Live Time

  • There is something wrong with cset not existing on the new hosts.
  • We also need to fix the renicing issue in
  • Jaroslav gave me permission to take a time profile on one of the farm hosts.
  • We need to increase the shared memory segment depth. It is currently set to 4.

SNEWS Trigger Test

  • Ron turned on the network retransmission metric in Ganglia. It was off because of the power outage.
  • Jan investigated a bit further to see how the retransmission rates looked with the larger triggers, but unfortunately found that we get the DL block errors at 2 seconds already from the BNEVB. He also saw the DL timeout with the full FD. In the end, the retransmission rates were small and not the source of our problem.
    • More details can be found here.
  • I might not have seen these errors because I was only using a partial detector (DB 5-11 only).
  • Jan and Ron performed some more network tests on Friday afternoon. When torturing datadisk-05, they saw write speeds above 1.2 GByte/s which is to be expected with a 10 Gbit/s link.

Trigger Rates

  • The trigger rates show a slight upturn since we have gained 10% of our live time by switching to the faster hosts.
  • Thanks to Kyle for uploading theses slides.