souravpaul8
commented
11 months ago My Code is as follows: `#include <cpprest/http_listener.h>
include <cpprest/json.h>
include <cpprest/http_client.h>
include "cpprest/uri.h"
include
include <boost/algorithm/string.hpp>
include "http_server.hh"
using namespace web; using namespace web::http; using namespace web::http::experimental::listener;
const int timeoutSeconds = 5; FILE *log_file;
const utility::string_t SERVER_URL = U("http://172.16.0.2:80"); const utility::string_t SERVICE_DISCOVERY_SERVER_URL = U("http://localhost:5001");
void getTimestamp(char timestamp, int timestampSize) { struct timeval tv; struct tm tm_info;
gettimeofday(&tv, NULL);
tm_info = localtime(&tv.tv_sec);
snprintf(timestamp, timestampSize, "%02d:%02d:%02d.%06ld",
tm_info->tm_hour, tm_info->tm_min, tm_info->tm_sec, tv.tv_usec);}
void handle_get(const http_request& request) { char timestampStart[20]; // Adjust the size as needed getTimestamp(timestampStart, sizeof(timestampStart));
utility::string_t path = request.request_uri().path();
std::vector<utility::string_t> segments;
boost::split(segments, path, boost::is_any_of("/"), boost::token_compress_on);
utility::string_t token = segments[1];
utility::string_t link;
for(int i = 2;i<segments.size();i++){
link += "/" + segments[i];
}
//std::cout << link << std::endl;
http::uri serviceDiscoveryUri = http::uri(U("http://172.16.0.4:5001"));
http::client::http_client serviceDiscovery(http::uri_builder(serviceDiscoveryUri)
.append_path(U("/redis/"))
.append_path(U(token))
.to_uri());
char timestampDiscoveryRequestStart[20]; // Adjust the size as needed
getTimestamp(timestampDiscoveryRequestStart, sizeof(timestampDiscoveryRequestStart));
serviceDiscovery.request(methods::GET)
.then([=](pplx::task<http_response> task)
{
http_response resp;
try {
resp = task.get();
}catch(std::exception &e) {
std::cout << "ServiceDiscovery: EXCEPTION CAUGHT: " << e.what() << std::endl;
}
char timestampDiscoveryRequestEnd[20]; // Adjust the size as needed
getTimestamp(timestampDiscoveryRequestEnd, sizeof(timestampDiscoveryRequestEnd));
string discovery_begin_time = resp.headers()["discovery_begin_time"];
string discovery_end_time = resp.headers()["discovery_end_time"];
resp.extract_json()
.then([=] (pplx::task<json::value> task)
{
json::value regResp;
string message;
string discoveryToken;
try {
regResp = task.get();
} catch (std::exception& e) {
std::cout << "ServiceDiscovery: EXCEPTION CAUGHT: " << e.what() << std::endl;
}
message = regResp["message"].as_string();
discoveryToken = regResp["tokenValue"].as_string();
http::uri redisServerUri = http::uri(U(message));
http::client::http_client redis(http::uri_builder(redisServerUri)
.append_path(U("/"))
.append_path(U(discoveryToken))
.to_uri());
char timestampRedisRequestStart[20]; // Adjust the size as needed
getTimestamp(timestampRedisRequestStart, sizeof(timestampRedisRequestStart));
redis.request(methods::GET)
.then([=](pplx::task<http_response> task)
{
http_response resp;
try {
resp = task.get();
}catch(std::exception &e) {
std::cout << "Redis: EXCEPTION CAUGHT: " << e.what() << std::endl;
}
char timestampRedisRequestEnd[20]; // Adjust the size as needed
getTimestamp(timestampRedisRequestEnd, sizeof(timestampRedisRequestEnd));
string redis_begin_time = resp.headers()["redis_begin_time"];
string redis_end_time = resp.headers()["redis_end_time"];
resp.extract_json()
.then([=] (pplx::task<json::value> task)
{
json::value regResp;
try {
regResp = task.get();
} catch (std::exception& e) {
std::cout << "Redis: EXCEPTION CAUGHT: " << e.what() << std::endl;
}
string message = regResp["Message"].as_string();
string redisToken = regResp["tokenValue"].as_string();
// std::cout << message << std::endl;
// if(redisToken != discoveryToken || redisToken != token || discoveryToken != token)
// std::cout << "Some issue with tokens" << std::endl;
http_response response;
HTTP_Response *html_response = handle_request(link);
if(html_response->status_code == "200")
response.set_status_code(status_codes::OK);
else
response.set_status_code(status_codes::NotFound);
response.headers().add(U("Content-Type"), U("text/html"));
response.set_body(utility::conversions::to_string_t(html_response->body));
response.headers().add(U("Date"),html_response->time);
response.headers().add(U("tokenValue"),token);
char timestampEnd[20];
getTimestamp(timestampEnd, sizeof(timestampEnd));
response.headers().add(U("mainserver_begin_time"),U(timestampStart));
response.headers().add(U("mainserver_end_time"),U(timestampEnd));
response.headers().add(U("mainserver_discovery_start_time"),U(timestampDiscoveryRequestStart));
response.headers().add(U("mainserver_discovery_end_time"),U(timestampDiscoveryRequestEnd));
response.headers().add(U("mainserver_redis_start_time"),U(timestampRedisRequestStart));
response.headers().add(U("mainserver_redis_end_time"),U(timestampRedisRequestEnd));
response.headers().add(U("discovery_begin_time"),U(discovery_begin_time));
response.headers().add(U("discovery_end_time"),U(discovery_end_time));
response.headers().add(U("redis_begin_time"),U(redis_begin_time));
response.headers().add(U("redis_end_time"),U(redis_end_time));
request.reply(response);
});
});
});
});}
int main() { http_listener listener(SERVER_URL);
listener.support(methods::GET, handle_get);
try {
listener.open().then([&listener]() {
std::cout << "Server started listening on " << SERVER_URL << std::endl;
}).wait();
std::cout << "Press Enter to exit." << std::endl;
std::cin.ignore();
listener.close().wait();
}
catch (const std::exception& e) {
std::cerr << "Server error: " << e.what() << std::endl;
}
return 0;} `
I have used cpprestsdk library to build multiple services running on different linux N/W namespaces. But I am facing an issue of periodic drop in throughput during my load test, when one of the service is pinned to 1 core of the machine and saturated to 100% utilization. But this doesn't happen when all the services are running in the same linux network namespace or localhost.
This is really not happening in the servers which uses socket programs instead of cpprestsdk. Can you help me on how to figure out the issue? Following are the time series throughput graph and perf profiling for the high throughput region (left) vs low throughput (right) region.