b6afacfc1e
The connection timeout was 2 minutes and the keep-alive timeout was 11 minutes. If a vector store node became unreachable, these long timeouts caused significant delays before the system could recover, negatively impacting high availability. This change aligns both timeouts with the `request_timeout` configuration, which defaults to 10 seconds. This allows for much faster failure detection and recovery, ensuring that unresponsive nodes are failed over from more quickly.
50 lines
1.3 KiB
C++
50 lines
1.3 KiB
C++
/*
|
|
* Copyright (C) 2025-present ScyllaDB
|
|
*/
|
|
|
|
/*
|
|
* SPDX-License-Identifier: LicenseRef-ScyllaDB-Source-Available-1.0
|
|
*/
|
|
|
|
#pragma once
|
|
|
|
#include <seastar/core/sstring.hh>
|
|
#include <seastar/core/temporary_buffer.hh>
|
|
#include <seastar/net/api.hh>
|
|
#include <vector>
|
|
|
|
namespace vector_search {
|
|
|
|
inline seastar::sstring response_content_to_sstring(const std::vector<seastar::temporary_buffer<char>>& buffers) {
|
|
seastar::sstring result;
|
|
for (const auto& buf : buffers) {
|
|
result.append(buf.get(), buf.size());
|
|
}
|
|
return result;
|
|
}
|
|
|
|
inline seastar::net::tcp_keepalive_params get_keepalive_parameters(std::chrono::milliseconds timeout) {
|
|
constexpr unsigned retry_count = 3;
|
|
constexpr auto min_interval = std::chrono::seconds(1);
|
|
|
|
// Divide total time by (count + 2) to balance Probing vs. Idle (~60/40)
|
|
auto target_duration = std::chrono::duration_cast<std::chrono::seconds>(timeout);
|
|
auto interval = target_duration / (retry_count + 2);
|
|
|
|
// Ensure minimum 1s interval
|
|
if (interval < min_interval) {
|
|
interval = min_interval;
|
|
}
|
|
|
|
auto idle = target_duration - (interval * retry_count);
|
|
|
|
return seastar::net::tcp_keepalive_params{
|
|
.idle = idle,
|
|
.interval = interval,
|
|
.count = retry_count,
|
|
};
|
|
}
|
|
|
|
|
|
} // namespace vector_search
|