Taraxa Weekly Tech Update: Week 50.

Resolving some PBFT block production and block proposal issues, more work on the social listening assistant.

Olga Grinina
Taraxa Project
2 min readDec 21, 2021

--

✅ 🚦If you have a question about why your node is not producing blocks, or producing the blocks slowly, please check out the F.A.Q. here: https://docs.taraxa.io/faq/testnet#what-makes-one-node-generate-more-blocks-than-another

Testnet node operations:

  • Debugged and investigated on the testnet is down issue, monitoring the PBFT progress in all of our consensus nodes.
  • Checked and investigated the binomial distribution for PBFT blocks proposal rate, and PBFT blocks production rate for all delegation players.
  • Added a binary search for the binomial distribution, in order to avoid nodes looping million times to verify another node's votes.
  • Added the maximum PBFT lambda value to be one hour for exponential backoff for advanced PBFT steps to avoid waiting too long in PBFT steps.
  • Add proposal committee size as 200 for PBFT proposals a maximum number of votes and blocks. Since PBFT will only pick up one winner block to vote, the regular committee size of 1000 is too much for proposal blocks.

Application:

We are now in the process of building out an open API to allow for integrations with popular messenger platforms to be able to elicit and track the trending data and sentiment: getting consistent data from Telegram groups and channels.

  1. Telegram Ingester:
  • Parsed combot groups to be able to send to the social listen assistant.
  • Worked on max retries error raise after many telegram_code failures.

2. Social-Listen-Assistant:

  • FTB-61 added new fields to assistant: last read message timestamp, number of the read messages in a group/chat, and a total message count.
  • FTB-62 endpoints sync between ingester and assistant: updated chat endpoint to receive data from the ingester.

--

--