Automatic peer dialing in bootstrap module after address resolution.

[ankur12-1610]

Related PR: #849

This PR adds Automatic peer dialing in bootstrap module after address resolution.

What was wrong?

Issue #798

Currently, the Bootstrap module in py-libp2p only resolves the multiaddrs of bootstrap peers and adds them to the peerstore. This enhancement proposes that, after resolving the addresses, the module should also attempt to connect (dial) to these peers automatically. This aligns with the intended purpose of bootstrapping in decentralized P2P networks, where nodes should proactively establish connectivity with known peers upon startup.

How was it fixed?

• After address resolution, swarm initiates connections to each bootstrap peer using trio nurseryconcurrent address processing
• Logs the outcome of connection attempts for observability
• Added IPv4+TCP address validation (currently supported transport)

Key Benefits

• The use of trio.open_nursery enables simultaneous connection attempts, leading to a significant reduction in bootstrap time
• Structured concurrency with proper error handling and cleanup
• Comprehensive connection attempt logging
• A trio.move_on_after function ensures that connection attempts are canceled if they exceed a set timeout

Implementation details

In order to test it properly first change examples/bootstrap/bootstrap.py:

by replacing

    try:
        async with host.run(listen_addrs=[listen_addr]):
            # Keep running and log peer discovery events
            await trio.sleep_forever()
    except KeyboardInterrupt:
        logger.info("👋 Shutting down...")

to

 try:
        async with host.run(listen_addrs=[listen_addr]):
            # Keep running and log peer discovery events
            while True:
                logger.info(f"Connected peers: {len(host.get_connected_peers())}")
                await trio.sleep(5)
            # await trio.sleep_forever()
    except KeyboardInterrupt:
        logger.info("👋 Shutting down...")

this helps you log the connected peers.

Now the core of the PR lies in using trio.open_nursery when calling _process_bootstrap_addr.

In essence, nursery.start_soon() transforms what would be a slow, sequential bootstrap process into a fast, parallel, and safe concurrent operation with proper error handling and resource management. This has significantly made the bootstrap process faster.

The core function which handles connection logic -

py-libp2p/libp2p/discovery/bootstrap/bootstrap.py

Line 191 in a8a71b0

async def _connect_to_peer(self, peer_id: ID) -> None:

Connection timeout feature is implemented using trio.move_on_after -

py-libp2p/libp2p/discovery/bootstrap/bootstrap.py

Line 219 in a8a71b0

with trio.move_on_after(self.connection_timeout):

Discussion Points

Future work could include adding a backoff logic to handle repeated connection failures, similar to the approach used in go-libp2p. This would further improve the resilience of the bootstrapping process.

Learning

Gaining a solid understanding of trio was a significant learning curve, and this wouldn't have been possible without the insightful reviews from @sumanjeet0012.

Please share your thoughts on the same and feel free to contribute.

[ankur12-1610]

@sumanjeet0012 is it apt to have a separate issue for updating the code of bootstrap example for logging connected peers?

[sumanjeet0012]

@ankur12-1610 The primary purpose of logging connected peers in the bootstrap example is to determine the time taken for bootstrapping and to identify which bootstrap nodes are successfully connected. Since the PR has already been merged, there is no need to address this now.

[ankur12-1610]

Makes sense.