2020-10-29 Meeting notes

Date

Attendees

Discussion items

WhoNotes
Stepan Lavrentev
  • There are two small problems now
    • WSV desync with the blockstore; probably after the node reboot. There is an idea of how to fix it on the development side.
    • There is a problem with the use of environment variables for configuration. Andrei Lebedev is helping with that.
  • Currently, I will try to reconfigure compose and try to start Iroha as is
  • Stepan will provide additional data on the issue with usage of environment variables
    • Stepan: Iroha Postgres host set among all configs equally. But with env vars that approach does not work.
    • Currently, it seems like an issue in the environment. I will take a look. 
  • The benchmark of Iroha operations is ready to present first results, now it covers "CreateRole" command.
    • RocksDB implementation is faster than Postgres 150x times.
    • I will proceed with further implementation of benchmark and RocksDB integration.
    • The transition from Postgres to RocksDB will be quite smooth, according to the results of research
    • I have an idea about transition implementation for all commands
  • The proposed solution for the WSV desync may help, but in general, there may be better ideas.
    • The failure of the peer happens during the process of applying blocks. However, because there were no logs, we should not say exactly.
    • Improvement of logging levels and better quality of the log handling will help us to investigate that problem and also do that in the future.
    • Stepan: the load for this issue to happen is following – intensive processing of MST transactions
    • We can keep the issue as "known" one for the 1.2.0 release, as it happens under heavy load and can be fixed by restarting with WSV recreation.
    • Stepan: I have tried to run the longevity stand without MST and there were no problems during the whole week of load testing.
    • There are some known issues in MST, which we need to fix. For example, not having an upper limit for MST storage.
  • Regarding Shepherd PR – the LinuxONE server cannot build the solution. Hence CI is not working correctly and we need to fix it on the LinuxONE.
    • I have contacted admins of our LinuxONE server, they said that server will soon go back to the online state.
    • Waiting for the Linux ONE to be configured.