Skip to end of metadata
Go to start of metadata

Date

Attendees

Local: Nate PeaseIgor GaponenkoVaikunth ThukralJohn Gates

Remote: Fabrice JammesBrian Van KlaverenAndy SalnikovFritz Mueller (for first half)

Discussion items

Fabrice:

  • Waiting on NCSA machines, sent requirements to them, preparing to use puppet there, implementing prototype for qserv
  • Some Swarm issues are fixed, xrootd starts now, but docker network doesn't work correctly. Some other features not available
  • Ok to keep working on it till PDAC is up
  • Moving data to OpenStack cluster for large scale CI, using ceph

 

AndyS:

  • No progress on supertask
  • Both AndyS and Gregory busy on other stuff
  • L1DB - writing some tests on the beefy in2p3 machine, preliminary tests show better performance than regular disk
  • Running on SSD and NVMe, looks similar so far

 

Brian:

  • Code of Conduct 95% established for LSSTC-Slack
  • Working on docker files for webserv etc
  • Working with Steve for recipe for docker @ NCSA
  • AndyH - mlock available only on new linux kernels, but container kernel is based on host OS, discussion needed?

 

Igor:

  • Jason Alt says PDAC is online, can log in to nodes
  • Ready to try to load stripe82 dataset
  • Chunk size needs to be decided, asked Serge and waiting on reply
  • Running some tests for settings
  • Data migration ~30% complete, ~2 weeks more for 100% (IN2P3 fraction 50%, NCSA 50%), these are calexps
  • Loading - considering partitioning first then using loader
  • EFD discussion with Fritz/Mario on Tuesday

 

John:

  • DM-6814 reviewed and being tested, needed updated xrootd with SID changes
  • Missing rows problem - hit czar with lots of queries, it returns fewer rows than expected. Will test and push to review after DM-6814 testing
  • System able to handle large load (cancel, wait etc) but only runs 4-5 FTS at a time
  • Malformed queries create executive and register it, causes segfault, made ticket for a quick fix
  • Some more issues that have been shelved for now
  • Retry mechanism will need to be reworked

 

Nate:

  • Composite datasets - implemented basic GET, now working on PUT and making unit tests
  • Once these are working, getting it reviewed and closing ticket
  • Look at further use/edge cases
  • All reqs dumped in confluence page
  • Jim Bosch to work the API

 

Vaikunth:

  • Found and worked around bug in elasticsearch query results for hosts
  • Have basic query throughput  (QTP) quantified
  • Running tests again after cluster restart - profile QTP against system params?

 

Other:

  • Setting up interview 2 with candidate 
  • xrootd bug for gcc version 6 or above fixed (DM-7679)