TL;DR

Running from source (Linux)

sed -i 's%const REQUEST_BLOCKS_PER_MESSAGE.*%const REQUEST_BLOCKS_PER_MESSAGE = 2%' ironfish/src/syncer.ts
yarn start:once start

Running from source (Windows)

• Open ironfish/src/syncer.ts in Notepad
• Edit const REQUEST_BLOCKS_PER_MESSAGE = 20 to const REQUEST_BLOCKS_PER_MESSAGE = 2
• Save the file
• yarn start:once start

Running from Docker (Linux and Windows)

❗This assumes you did NOT run your container with --rm else the changes are gone on restart!!❗

I expect your container to be run with: docker run -d --name ironfish --network host ghcr.io/iron-fish/ironfish:latest ........

Find NAME_OF_YOUR_DOCKER_CONTAINER with docker ps -a

docker exec -t NAME_OF_YOUR_DOCKER_CONTAINER sh -c "sed -i 's%REQUEST_BLOCKS_PER_MESSAGE.*%REQUEST_BLOCKS_PER_MESSAGE = 2%' /usr/src/app/node_modules/ironfish/build/src/syncer.js"
docker restart NAME_OF_YOUR_DOCKER_CONTAINER
docker logs -f NAME_OF_YOUR_DOCKER_CONTAINER

---

Problem

A lot of issues on GH and Discord about peers disconnecting while syncing. I’ve done some debugging:

Reproduction of the bug:

My initial start: https://github.com/iron-fish/ironfish/blob/master/ironfish/src/syncer.ts#L358-L362

      this.logger.info(
        `Requesting ${this.blocksPerMessage} blocks starting at ${HashUtils.renderHash(
          head,
        )} (${sequence}) from ${peer.displayName}`,
      )

I noticed blocksPerMessage is 20, being passed in the ctor: https://github.com/iron-fish/ironfish/blob/master/ironfish/src/syncer.ts#L65

My hunch was the peers are timing out sending 20 blocks, so I lowered it to 1:

Upon restarting, this solved the problem and my chain is syncing again:

Experimenting with a number, I found 10 is still to high, but 5 works.

    this.blocksPerMessage = 5; //options.blocksPerMessage ?? REQUEST_BLOCKS_PER_MESSAGE

Short term fix is to use a lower blocksPerMessage, long term fix is to make sure peers dont disconnect when sending blocks over.

I can create the PR for the short term fix if needed.