Fix WSAEALREADY (0X00002735): Nonblocking Socket Error

1. The Connection Timeout Problem

The most common trigger for WSAEALREADY is when your app tries to connect a nonblocking socket and doesn't wait long enough before retrying. I see this all the time in custom chat clients, IoT devices, and server monitors. Had a client last month whose inventory management app locked up because it kept hammering a socket with new connect() calls after the first one hadn't finished.

When you call connect() on a nonblocking socket, it returns immediately with WSAEWOULDBLOCK (10035). That means the connection is in progress. If you then call connect() again on that same socket before the first attempt completes, you get WSAEALREADY. The socket is already working on it.

The fix is simple: track the socket state and don't retry the connect until you get a response. Use select(), poll(), or WSAAsyncSelect() to wait for completion. Check for FD_CONNECT events. If you're using overlapped sockets, wait on the event handle.

// C++ example of proper nonblocking connect handling
SOCKET s = socket(AF_INET, SOCK_STREAM, 0);
u_long mode = 1;
ioctlsocket(s, FIONBIO, &mode);

struct sockaddr_in addr;
addr.sin_family = AF_INET;
addr.sin_port = htons(8080);
inet_pton(AF_INET, "192.168.1.100", &addr.sin_addr);

// Only attempt connect once
int result = connect(s, (struct sockaddr*)&addr, sizeof(addr));
if (result == SOCKET_ERROR && WSAGetLastError() != WSAEWOULDBLOCK) {
    // handle error
} else {
    // Wait with select() or WSASelect()
    fd_set writeSet;
    FD_ZERO(&writeSet);
    FD_SET(s, &writeSet);
    struct timeval tv = {5, 0}; // 5 second timeout
    if (select(0, NULL, &writeSet, NULL, &tv) > 0) {
        // Connection succeeded (check getsockopt for errors)
    }
}

Set a reasonable timeout. Don't use zero—you'll spin in a loop. I usually set 5-10 seconds for local networks, 30 for internet connections.

2. Retry Logic Spinning Out of Control

The second cause is retry loops that don't check error codes properly. Say you have a loop that calls connect() until it succeeds or a counter expires. If you don't filter WSAEALREADY, your app will keep burning CPU calling connect() millions of times per second. I fixed a point-of-sale system once where this caused the whole terminal to freeze.

The pattern is always: application gets WSAEWOULDBLOCK, ignores it, calls connect again, gets WSAEALREADY. Then it either ignores that too and loops forever, or misinterprets it as a permanent failure and shuts down.

Fix it by:

Using a state machine for socket operations. States: IDLE, CONNECTING, CONNECTED, ERROR.
After connect() returns WSAEWOULDBLOCK, set state to CONNECTING and wait for an event using select() or WSAEventSelect().
Never call connect() again on a socket in CONNECTING state unless you close it first.
If you must retry, close the old socket and create a new one. Don't reuse.

// Bad retry loop — causes WSAEALREADY
while (retries < 5) {
    if (connect(s, ...) == SOCKET_ERROR) {
        int err = WSAGetLastError();
        if (err == WSAEALREADY) continue; // loops forever
    }
    retries++;
}

Replace with a proper async wait or timer-based retry.

3. Bind Conflicts on the Same Port

Less common, but I've seen it in services that bind a socket to a specific port, then try to connect from that same port without waiting. If you call bind() on a nonblocking socket and then immediately connect(), you might get WSAEALREADY if the bind operation hasn't completed yet. Winsock does allow bind on nonblocking sockets, but it's not guaranteed to be instant.

This happens in some homebrew VPN or proxy software that tries to re-use the same source port. The socket is still in the process of binding when the connect call hits.

Fix: Don't set the socket nonblocking before bind. Bind first, then set to nonblocking mode. Or if you need nonblocking bind, wait for FD_CONNECT or FD_ACCEPT events before proceeding.

// Correct order: bind first, then set nonblocking
SOCKET s = socket(AF_INET, SOCK_STREAM, 0);
struct sockaddr_in local;
local.sin_family = AF_INET;
local.sin_port = htons(5000);
local.sin_addr.s_addr = INADDR_ANY;
bind(s, (struct sockaddr*)&local, sizeof(local));

// Now set nonblocking
u_long mode = 1;
ioctlsocket(s, FIONBIO, &mode);

Also check if you have another socket already bound to that port. Use netstat -ano | findstr 5000 to see conflicts.

Quick-Reference Summary

Cause	What's Happening	Fix
Connection timeout	connect() called while previous connect still in progress	Use select() to wait for completion; don't retry early
Retry logic error	Loop calling connect() without checking WSAEALREADY	State machine; close and recreate socket on retry
Bind conflict	bind() not finished before connect() on nonblocking socket	Bind first, then set nonblocking; avoid port reuse

If none of these work, check your firewall and antivirus—I've seen McAfee and Norton block socket operations and cause WSAEALREADY errors. Disable those temporarily to test.