Skip to end of metadata
Go to start of metadata

Date

May 15, 2014

Attendees

Discussion Items

Greg:

 * Created issues, will space out across sprints and add estimates

 * One of them is to follow up with HTCondor team on dynamic slot ClassAds

 * Adding visit id to worker job for monitoring

 * Working on pre-script to write DAG dynamically

   + Will keep simple; probably not so necessary for Nightly Computing

Steve:

 * Will estimate story points

 * Have replicator jobs talking to replicator nodes and distributor nodes

   + Just metadata for now, not data yet

   + How much data? Will be 500+MB; maybe use 50MB scaled to network

   + How to transfer data?

     - Job to node: shm, socket

     - Replicator to distributor: memory to memory via socket, file transfer

     - Doesn't matter much; whatever is easiest

 * Using regular lsst-dev HTCondor pool

 * Right now, some scripts point to SRP's home directory

   + Will work on making relocatable

 

OCS will have DDS interface available sooner than expected, maybe June/July

DonP:

 * Log messages gathered to one location? Yes

   + Also critical messages can go via real-time messages

   + Also status updated directly in databases for monitoring

 * Need to make sure operators have sufficient insight into system

 * Is Tucson headquarters an overall control center?

   + Anticipate HQ is mostly overall management, not day-to-day

   + Four operational centers (Base, Archive, Chile DAC, US DAC)

   + Operational centers run mostly independently

   + Base runs Nightly Computing in conjunction with NCSA

   + Archive runs DRP

 * DES is trying out large solid-state disk behind GPFS

   + K-T will see if there are any potential uses for LSST testing

   + Maybe DB group?

Action Items

  • Kian-Tat Lim will see if there are any potential uses of large solid-state disk behind GPFS for LSST testing