| Who | Notes |
|---|
| - We need Andrei to discuss out backup-related improvements
- iOS library is under development, we need to continue to manage Denis and his progress
- Community bot is going to be updated
|
| - Had a meeting with Bulat about the preparation for performance testing.
- Currently "time to commit" metric is broken
- Discussed also "blockstore height" metric collection
- Locust is now working for 10 minutes only and then stops making new requests.
- Currently, I am looking to the issue reasons. Seems like that it is because of RPC connection timeout.
- After fixing all issues and running on Longevity, we will start a dedicated Performance testing environment and start testing on it.
- I will do that today or tomorrow.
- Salakhiev Kamil : Do you implemented a selection of several modes of testing?
- Not yet, will check and implement it.
|
| - Working on "Add peer" hands-on operation preparation for the documentation
- Need support in Python script for that operation. We need to create a script which will add peers to the network.
- Will try to work with Mattermost-based bridge bot to get understanding about its qualities
|
| - Updated approach to the peer connection without the overload of the network. Need to consolidate it with Andrey, as there are two different ways and solutions.
- Regarding the GRPC problem: I have understood the flow of all threads.
- Mikhail Boldyrev answered many questions about the project architecture – respect!
- There are several important specifics:
- In consensus, there are some situations when the round should be rejected, but actually, it is not.
- There is a timer for resending votes. It works every second. But for the "poor connection" nodes, it cannot happen, and it leads to the infinite retries of finding consensus.
- It can happen, that there can be different proposals from the same ordering service. Need to investigate the code.
- Seems like that we need to add the additional mechanism in consensus algorithm to handle such situations. We need to reject proposals in all the situations when there are more than f malicious nodes.
|