Hi,
I was making some tests with the Echo container, including ensuring connection pool logics / configuration.
Currently this is taking the default 5s Node keepAliveTimeout which will then start dropping connections. If we would be able to stablish the keepAliveTimeout we would be able to allow the consumer side to manage the connections.
I've hacked a solution as:
// How long an idle keep-alive connection is held before the server closes it.
// Node's default is 5s, so an upstream keep-alive pool in front of this server can never be observed intact.
// Raising it past the drain time lets the pool settle and leaves the reverse proxy in charge of trimming it.
const keepAliveTimeoutMs = parseInt(process.env.KEEPALIVE_TIMEOUT_MS, 10);
if (Number.isFinite(keepAliveTimeoutMs) && keepAliveTimeoutMs > 0) {
for (const server of [httpServer, httpsServer]) {
server.keepAliveTimeout = keepAliveTimeoutMs;
// headersTimeout bounds how long a started request may take to send its headers;
// keeping it above keepAliveTimeout stops Node from closing a pooled connection just as it is being reused.
server.headersTimeout = keepAliveTimeoutMs + 5000;
}
console.log(`Keep-alive timeout set to ${keepAliveTimeoutMs}ms (headers timeout ${keepAliveTimeoutMs + 5000}ms).`);
}
Hi,
I was making some tests with the Echo container, including ensuring connection pool logics / configuration.
Currently this is taking the default 5s Node keepAliveTimeout which will then start dropping connections. If we would be able to stablish the keepAliveTimeout we would be able to allow the consumer side to manage the connections.
I've hacked a solution as: