feat: lift the client's concurrency ceiling from ~15 peers to 1024

The client had never been driven at scale. A stress harness that mixes
searches, downloads, uploads and browses against a local soulfind found four
ceilings, three of which lose work silently rather than merely being slow.

Peers were capped at one per core. ActorSystem ran each actor's whole lifetime
as one job on a fixed-size ThreadPool, and actor loops block until stopped, so
the pool acted as a semaphore: past the cap an actor never started, never
connected and never errored, and uploads stopped working entirely. The pool had
exactly one consumer and every job it received was non-terminating, so it is
gone; actors get their own thread.

TransferResponse raced the bookkeeping it depends on. It was sent straight back
to the peer while UpdateDownloadTokens was still queued for the client loop —
but the response is what invites the peer's file connection, and that connection
is matched by token. Under load the client loop lost, the file connection
matched nothing, and the download sat on Queued forever. The response now goes
out after the token is recorded.

Download tokens were md5(filename) truncated to 20 bits, and the store deletes
every entry matching a token, so two downloads sharing one destroyed each other
— which fetching the same filename from two peers does every time. They now
come from a counter.

Port 0 from GetPeerAddress means the server does not know where a user listens.
It was cached and dialled, and the upload waiting on it was dropped. It is no
longer cached or served against; the connect attempt stays, because its failure
is what makes an unreachable peer fall back to server brokering.

Mixed profile at 1024 peers: 512/512 downloads, 256/256 uploads, 400-850 MiB/s
in ~7s, against 71/96 downloads at 7.9 MiB/s and a 60s timeout before.
This commit is contained in:
Michel de Graaf
2026-07-26 17:01:32 +02:00
parent 38a231ffd1
commit 5b4ef30f08
12 changed files with 1539 additions and 191 deletions
+41 -5
View File
@@ -316,9 +316,26 @@ impl Client {
obfuscated_port
);
// Cache the address for the serve/search paths,
// and collect any uploads waiting on it.
let waiting_serves =
// Port 0 is the server saying it does not know
// where this user listens (no SetWaitPort yet,
// or firewalled). It is not an address: caching
// it poisons every later lookup, and an upload
// dialled at it dies with "Can't assign
// requested address" and is dropped on the
// floor. Leave those uploads queued for a later
// resolution instead.
//
// The connect attempt further down still runs:
// its failure is exactly what makes an
// unreachable peer fall back to the
// server-brokered path.
let waiting_serves = if port == 0 {
warn!(
"[client] server reports no listening port for {}",
username
);
Vec::new()
} else {
match client_context.write_safe() {
Ok(mut ctx) => {
ctx.cache_peer_address(
@@ -331,7 +348,8 @@ impl Client {
.unwrap_or_default()
}
Err(_) => Vec::new(),
};
}
};
for token in waiting_serves {
Self::spawn_serve(
&client_context,
@@ -432,7 +450,7 @@ impl Client {
context.add_download(Download {
username: username.clone(),
filename: transfer.filename,
filename: transfer.filename.clone(),
token: transfer.token,
size: transfer.size,
download_directory: download
@@ -444,6 +462,24 @@ impl Client {
});
context.remove_download(old_token);
}
// Only now invite the file connection: it is
// matched by this token, which is recorded as
// of the line above. Answering any earlier
// races the peer's connection against our own
// bookkeeping.
let registry = context.peer_registry.clone();
drop(context);
if let Some(registry) = registry {
let response =
crate::message::server::MessageFactory::build_transfer_response_message(
transfer,
);
let _ = registry.send_to_peer(
&username,
PeerMessage::SendMessage(response),
);
}
}
ClientOperation::UploadFailed(
username,