Overall this MR seems to greatly improve the initial sync and steady state of electrs alongside a slower bitcoind.

some issues found during testing

• the electrum method blockchain.transaction.get will panic on the handling thread if the transaction is not known by electrs (panic starts at https://github.com/shesek/electrs/blob/202405-mempool-sync/src/electrum/server.rs#L375
• some requests from electrs to bitcoind will disconnect waiting for recv, at https://github.com/shesek/electrs/blob/202405-mempool-sync/src/daemon.rs#L210
  Increasing the rpcthreads parameter in bitcoind does not seem to change this. Electrs itself seems to recover. These recv disconnects happen with different rpcs (usually getrawtransaction, but also getbestblockhash)

* some requests from electrs to bitcoind will disconnect waiting for recv, at https://github.com/shesek/electrs/blob/202405-mempool-sync/src/daemon.rs#L210
  Increasing the `rpcthreads` parameter in bitcoind does not seem to change this. Electrs itself seems to recover. These recv disconnects happen with different rpcs (usually `getrawtransaction`, but also `getbestblockhash`)

While tracing this with wireshark, I observed that electrs sends a request (such as getbestblockhash), and waits for a response. After 30 seconds (and a few TCP Keep-Alive interactions), bitcoind sends a FIN,ACK without any text response. Seems to be some behaviour in bitcoind.

the electrum method blockchain.transaction.get will panic on the handling thread if the transaction is not known by electrs

I was unable to reproduce this, the "missing transaction" message appears to be returned as a JSONRPC error without panicking:

# Client
$ echo '{"id": 1, "method": "blockchain.transaction.get", "params": ["0000000000000000000000000000000000000000000000000000000000000000"]}' | nc 172.17.0.3 60401
{"error":"missing transaction","id":1,"jsonrpc":"2.0"}

# Server logs
TRACE RPC Request("{\"id\": 1, \"method\": \"blockchain.transaction.get\", \"params\": [\"0000000000000000000000000000000000000000000000000000000000000000\"]}\n")
WARN rpc #1 blockchain.transaction.get [String("0000000000000000000000000000000000000000000000000000000000000000")] failed: Error: missing transaction
TRACE RPC Done
# ... keeps running

Seems to be some behaviour in bitcoind.

I'm seeing that behaviour on my local regtest env too.

utACK a6c181e

in analyzing pcap trace between local electrs and local. bitcoin rpc, we see electrs asking for getrawmempool on
3 times in 16 seconds, in a single socket connection. MR needs more analysis/testing.

2 issues

• repeated requests for full mempool on one thread (correct?)

• disconnects are actually coming from electrs, not bitcoind (screenshot of pcap)

following one of the "disconnects":

1. electrs makes RPC request for "getrawtransaction"
2. bitcoind responds with data
3. electrs sits on connection for 30 seconds
4. bitcoin sends FIN,ACK but electrs does not close
5. electrs attempts to send more
6. bitcoin sends RST

electrs should be responding to the FIN,ACK and active, and reopen

• we are also seeing 5 second delays after receiving 500 from bitcoin. The 500 error is used by bitcoin rpc to indicate that the getrawtransaction call requested a non existent tx, which will happen after blocks are mined, and in presence of full rbf. Electrs should consider this case of error by inspecting the details of the json-rpc response, and not wait 5 seconds.

repeated requests for full mempool on one thread ... we see electrs asking for getrawmempool on 3 times in 16 seconds, in a single socket connection

getrawmempool is used to get the list of mempool txids (not txs), and should be sent from the main (sync loop) thread using the same socket (unless it disconnects). It should be sent at least once every 5 seconds, possibly more if additional roundtrips are needed to get a full mempool view.

The 500 error is used by bitcoin rpc to indicate that the getrawtransaction call requested a non existent tx ... Electrs should consider this case of error by inspecting the details of the json-rpc response, and not wait 5 seconds.

This should already be the case. In fact only errors explicitly constructed as an ErrorKind::Connection would trigger the automatic reconnection behavior, other errors will bubble up. RPC errors for "tx not found" are identified and ignored, while other unexpected RPC errors terminate execution (without reconnecting).

disconnects are actually coming from electrs, not bitcoind (screenshot of pcap)

Looking into this.

electrs sits on connection for 30 seconds .. bitcoin sends FIN,ACK but electrs does not close

bitcoind appears to close the connection after a period of inactivity, which can be adjusted with the (hidden) config option rpcservertimeout (defaults to 30 seconds).

Adjusting it upwards made the disconnection issues I was experiencing in my local env go away. I also tested with electrs and bitcoind connected over the internet (local electrs connecting to a regtest bitcoind on a vps) and was able to retain open TCP connections with no disconnections for several hours (used every ~2 minutes, with rpcservertimeout set to 5 minutes).

I also tried investigating what's causing electrs not to close immediately in reply to bitcoind's FIN, but do not have an answer yet.

Rebased and added some more verbose logging for mempool syncing.

testing latest commit using local electrs and bitcoind, I set parameters in bitcoin.conf

rpcthreads=16
rpcservertimeout=600
rpcworkqueue=50

I have not seen any "disconnected from daemon while receiving" messages.

utACK 7a068bf