2020-12-14 Meeting notes

Date

Attendees

Discussion items

WhoNotes

Ruslan Ganeev

  • Updated scripts for performance testing
    • Used correct Locust methods and added required commands.
    • Currently cannot check them, as there are some issues on the local environment.
    • Want to do all settings through the UI, for better possibilities of on-site configuration
    • Will have a call with Bulat to define answers on all integration-related questions
    • Had some issues with Locust and block listener. Finally, did it through the single user.
    • We will have web UI with graphics if everything will go well.
    • Will finish today with all preparations.
  • Alexander will join Iroha from today on a full-time basis.
  • There is some small issue on the previous project, will finish it today or tomorrow.
  • Today checked logs from Konrad, there are some interesting points
    • Log synchronization is not going well. Sometimes two entries are overlapping. In the worst case, there is a race condition.
    • Found the lost thread on consensus (for more than 1 minute) – look like that it was the thread for the lost node. Seems like that there is a block.
    • On the GRPC call, there is a break in the thread for 16 seconds. Seems like that some calls are breaking down, we need to create a network with lost packets between them to get the behaviour and change in flow in such situation.
  • Today will perform the experiment with the lost packets. Will build and debug Iroha, with the watchdog thread.
    • Logs look like that we have unexpected locks in our threads.
    • Kamil Salakhiev : Do we have some kind of threading control mechanism?
    • GRPC is not limited by thread number, so it creates new ones when needed.
    • We can do parallel logic, but we should keep the synchronized state.