Attendees
- Brian
- Derek
- John
- Soundar
- Shawn
Updates
ESnet data
- Have code to send a day’s worth of data, but it’s sending too many records (28M) at once
- Will update to limit the flow rate of data
- Discussion on collector architecture
- Example checkpoint management code
- Shawn wrote up the Purdue topology case
- Record with OBS Studio?
perfSONAR to bus
Tape replay
- Now sending Feb 2019
- Added replay overview writeup to website
- Add a messages per second graph
Ongoing topics
XRootD TCP statistics
- Andy wrote plugin infrastructure, merged to master
- Hook will track disconnects only at this point
- Derek will develop plugin (shared library)
- Will send jstream (JSON) packets rather than binary
ESnet data
- Collector on github
- Uses Netbeam API
- Have access to SNMP (over 9022 interfaces)
- Data: traffic (2 years); errors, discards (1 month)
- Size: About 1GB/mo
- Working on reviewing schema and moving into production
- Don’t send to tape now. Wait until we’re ready to move to production.
Sending HTCondor data to both ES instances
- Chicago ES is newer, and plans to use newer features (ILM)
- This incompatible with ES on the older GRACC instance
- Should the data go to tape? Est. around couple hundred MB/day
- U Waterloo writeup is complete
- Purdue/FNAL, Madison, and throughput writeup is in progress
- Make recorded sessions
- Convert into webpage, add to SAND website (John)
HTCondor monitoring and emails
- Derek will send weekly emails to SAND team list