7/22/2019 Agenda & Notes

Owned by Mark Ford
Last updated: Jul 22, 2019
3 min read
  • Review of Sawtooth and v1.2 Status on Multi-Project Kanban Board (Mark)
    • Multiple groups working on LR testing, still in progress and release won’t happen until this is completed. Once LR tests have passed, release date shall be announced (including supporting blogs, etc). Doc updates will happen as part of the release. (Mark)
      • PoET is cleared on the Intel side (Dan M)
      • Bitwise hasn’t seen PoET LR7 pass and it’s testing might be more rigorous. Goal is to determine if there are regressions and if it is stable, rather than optimizing settings to get it to pass. Comparing 1.1 and 1.2 performance, with the 1.1 test passing (with PoET engine patch). 1.2 test still underway. Data still being analyzed at the time of this meeting (Shawn)
        • Ok with using recommended settings. Might not be worth rigorous testing because Transact will change things. (Dan M )
      • Noticed a performance hit in 1.2. Issue seems to be happening upon duplicate block arrival. (Arun)
      • Scheduler or another component causing forking, not picking up blocks from pending queue. (Amol)
      • Bitwise team going to do root cause analysis over the data collected over the weekend. The issues previously discussed could be the problem, but we will know more later in the week. (Shawn)
        • Not advocating for PoET performance optimization to hold up release. (Dan)
          • It isn’t performance, but possibly a more serious stability issue (Peter)
      • How far out is transact integration? (Dan)
        • Not for a while, probably not relevant for this issue (Shawn)
        • Probably a 2.0 feature (James)
    • Darian described the Simplified SDE/State epic and associated work.
  • Discuss PoET LR status, latest PR changes (Arun)
    • Passed PoET simulator and LR7 at 10 nodes passed
  • Validator sends duplicate block to the consensus engine (Arun)
    • Debug session requested. Issue wityh Logs not matching up with code
  • Discuss non-functional Education Simple Supply app (Dan A)
    • Course currently not functioning due to protobuf issue. Provided a few PRs to update proxies and update Sawtooth to 1.2.
    • The course and material should both be updated. No one has signed up to update it, so it doesn’t make sense to only update the code. (Shawn)
    • The raw course material should be made available (Dan A)
    • Other currently engaged maintainers should be added to the Sawtooth Simple Supply project since Darian and Anne Chennette are on other projects. (Darian)
  • Production stability of Sawtooth networks (Jon Geater)
    • Seth docker-compose and main docker compose are slightly different and hard to orchestrate together. 
    • Consistent Dockerfiles across projects have been talked about but not implemented
    • RFC for consistent Docker stuff might be an option. Will look at writing that RFC.
    • Do we want to have a recipe book with recommended settings etc?
    • Expected behaviour on restarts (aka 'fragility')
      • After restarting a node 3-4 times it will never catch up. Peer connection list grows and shrinks
        • Devmode is unreliable, highly simplistic and may cause issues if you are sending a lot of batches. PBFT might be a better option to test with. (Peter)
      • Raft eventually caught up after 4000 blocks. Trying to restart multiple times, connection list is not cleared out. (Arun)
  • Open RFCs
  • PR Status Discussions
    • PR 23 can be closed. (Arun)
    • Other PR statuses unclear at time of meeting due to Github server issues 
  • Open Forum
    • Discussed the pending queue bug - log review (Arun)
      • Sounds like a race condition. Could also be incomplete knowledge of state machine block model. Debugging would be fairly complex, might have to instrument code. (James)
      • Continued the debugging of this item on RocketChat