为什么向基于 epoll 的网络爬虫添加 MySQL 代码会减慢它的速度？
Why would adding MySQL code to an epoll based web crawler slow it down so much?
我有一个使用 libcurl 和 epoll 来抓取网站的工作代码：
https://github.com/JamesRead5737/libcurlmemoryleak/blob/master/crawler.c
典型输出为：
Parsed sites: 0, 1024 parallel connections, 10989 still running, 10989 transfers Exiting normally.
Parsed sites: 0, 0 parallel connections, 0 still running, 0 transfersrsrFinished all in progress downloads.
Exiting.
如您所见，代码达到了 1024 个并行连接的硬编码限制。这在我的专用服务器上最多消耗 3 Gbps。
像这样添加 Mysql 代码：
#include <errno.h>
#include <fcntl.h>
#include <signal.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/epoll.h>
#include <sys/stat.h>
#include <sys/time.h>
#include <sys/timerfd.h>
#include <sys/types.h>
#include <sys/resource.h>
#include <sys/socket.h>
#include <arpa/inet.h>
#include <time.h>
#include <unistd.h>
#include <pthread.h>
#include <netdb.h>
#include <time.h>
#include <netinet/in.h>
#include <mysql.h>
#include <curl/curl.h>
#define MSG_OUT stdout
#define DEFAULT_QUEUE_LENGTH 10000
#define mycase(code) \
case code: s = __STRING(code)
#define MAX_CONNECTIONS 1024
MYSQL *mysql_con;
/* Global information, common to all connections */
typedef struct _GlobalInfo
{
int epfd; /* epoll filedescriptor */
int tfd; /* timer filedescriptor */
CURLM *multi;
int still_running;
pthread_mutex_t lock;
int concurrent_connections;
pthread_mutex_t parsed_lock;
int parsed_sites;
int transfers;
} GlobalInfo;
int new_body_conn(char *url, GlobalInfo *g);
/* Information associated with a specific easy handle */
typedef struct _ConnInfo
{
CURL *easy;
char *url;
GlobalInfo *global;
char error[CURL_ERROR_SIZE];
size_t size;
char *data;
} ConnInfo;
/* Information associated with a specific socket */
typedef struct _SockInfo
{
curl_socket_t sockfd;
CURL *easy;
int action;
long timeout;
GlobalInfo *global;
} SockInfo;
void
mysql_stop()
{
mysql_close(mysql_con);
}
void
mysql_start()
{
mysql_con = mysql_init(NULL);
if (mysql_con == NULL)
{
fprintf(stderr, "%s\n", mysql_error(mysql_con));
exit(1);
}
if (mysql_real_connect(mysql_con, "localhost", "crawler", "password", "crawl", 0, NULL, 0) == NULL)
{
fprintf(stderr, "%s\n", mysql_error(mysql_con));
exit(EXIT_FAILURE);
}
if (mysql_query(mysql_con, "CREATE TABLE IF NOT EXISTS `frontier` (`url` varchar(2084) NOT NULL, `id` int NOT NULL AUTO_INCREMENT, PRIMARY KEY (`id`), UNIQUE KEY `url` (`url`), KEY `url_2` (`url`)) ENGINE=InnoDB AUTO_INCREMENT=1"))
{
fprintf(stderr, "%s\n", mysql_error(mysql_con));
mysql_stop();
exit(1);
}
if (mysql_query(mysql_con, "CREATE TABLE IF NOT EXISTS `crawled` (`id` int NOT NULL AUTO_INCREMENT, `url` varchar(2084) DEFAULT NULL, `title` varchar(768) DEFAULT NULL, `date` varchar(128) DEFAULT NULL, `last_modified` varchar(128) DEFAULT NULL, `links` int DEFAULT NULL, `backlinks` int DEFAULT NULL, `http_code` int DEFAULT NULL, PRIMARY KEY (`id`), UNIQUE KEY `url` (`url`), KEY `http_code` (`http_code`), KEY `title` (`title`)) ENGINE=InnoDB AUTO_INCREMENT=1"))
{
fprintf(stderr, "%s\n", mysql_error(mysql_con));
mysql_stop();
exit(1);
}
if (mysql_query(mysql_con, "CREATE TABLE IF NOT EXISTS `emails` (`email` varchar(2084) NOT NULL, `id` int NOT NULL AUTO_INCREMENT, PRIMARY KEY (`id`), UNIQUE KEY `email` (`email`), KEY `email_2` (`email`)) ENGINE=InnoDB AUTO_INCREMENT=737 DEFAULT CHARSET=latin1"))
{
fprintf(stderr, "%s\n", mysql_error(mysql_con));
mysql_stop();
exit(1);
}
}
void
mysql_url_visited_push(char *url, char *title)
{
char sql[8192];
char escaped_url[(strlen(url)*2)+1];
if (!mysql_real_escape_string(mysql_con, escaped_url, url, strlen(url)))
{
}
if (title != NULL)
{
char escaped_title[(strlen(title)*2)+1];
if (!mysql_real_escape_string(mysql_con, escaped_title, title, strlen(title)))
{
}
sprintf(sql, "INSERT INTO crawled (url, title) VALUES ('%s', '%s')", escaped_url, escaped_title);
} else {
sprintf(sql, "INSERT INTO crawled (url, title) VALUES ('%s', '%s')", escaped_url, title);
}
if (mysql_query(mysql_con, sql))
{
fprintf(stderr, "mysql_url_visited_push sql=%s %s\n", sql, mysql_error(mysql_con));
}
}
int
mysql_url_visited_find(char *url)
{
char sql[8192];
char escaped_url[(strlen(url)*2)+1];
if (!mysql_real_escape_string(mysql_con, escaped_url, url, strlen(url)))
{
}
sprintf(sql, "SELECT * FROM crawled WHERE url = '%s'", escaped_url);
if (mysql_query(mysql_con, sql))
{
fprintf(stderr, "mysql_url_visited_find sql=%s %s\n", sql, mysql_error(mysql_con));
} else {
MYSQL_RES *result = mysql_store_result(mysql_con);
if (!result)
{
fprintf(stderr, "%s\n", mysql_error(mysql_con));
} else if (result > 0){
mysql_free_result(result);
return 1;
} else if (result == 0) {
mysql_free_result(result);
return 0;
}
}
}
int
mysql_url_frontier_find(char *url)
{
char sql[8192];
char escaped_url[(strlen(url)*2)+1];
if (!mysql_real_escape_string(mysql_con, escaped_url, url, strlen(url)))
{
}
sprintf(sql, "SELECT * FROM frontier WHERE url = '%s'", escaped_url);
if (mysql_query(mysql_con, sql))
{
fprintf(stderr, "mysql_url_frontier_find sql=%s %s\n", sql, mysql_error(mysql_con));
} else {
MYSQL_RES *result = mysql_store_result(mysql_con);
if (!result)
{
fprintf(stderr, "%s\n", mysql_error(mysql_con));
} else if (result > 0){
mysql_free_result(result);
return 1;
} else if (result == 0) {
mysql_free_result(result);
return 0;
}
}
}
void
mysql_url_frontier_push(char *url)
{
char sql[8192];
char escaped_url[(strlen(url)*2)+1];
if (!mysql_real_escape_string(mysql_con, escaped_url, url, strlen(url)))
{
}
sprintf(sql, "INSERT IGNORE INTO frontier (url) VALUES ('%s')", escaped_url);
if (mysql_query(mysql_con, sql))
{
fprintf(stderr, "mysql_url_frontier_push sql=%s %s\n", sql, mysql_error(mysql_con));
}
}
char *
mysql_url_frontier_pop()
{
char *url = NULL;
char sql[8192];
if (mysql_query(mysql_con, "SELECT url FROM frontier ORDER BY id") == 0)
{
MYSQL_ROW row;
MYSQL_RES *result = mysql_store_result(mysql_con);
if (result == NULL)
{
fprintf(stderr, "mysql_url_frontier_pop mysql_store_result sql=%s %s\n", sql, mysql_error(mysql_con));
exit(EXIT_FAILURE);
}
if ((row = mysql_fetch_row(result)))
{
url = strdup(row[0]);
char escaped_url[(strlen(url)*2)+1];
if (!mysql_real_escape_string(mysql_con, escaped_url, url, strlen(url)))
{
}
sprintf( sql, "DELETE FROM frontier WHERE url = '%s'", escaped_url);
if (mysql_query(mysql_con, sql))
{
fprintf(stderr, "mysql_url_frontier_pop mysql_query sql=%s %s\n", sql, mysql_error(mysql_con));
exit(EXIT_FAILURE);
}
}
mysql_free_result(result);
}
return url;
}
int
starts_with(const char *str, const char *pre)
{
size_t lenstr;
size_t lenpre;
if (str == NULL || pre == NULL)
return (-1);
lenstr = strlen(str);
lenpre = strlen(pre);
if (lenstr < lenpre)
return (-1);
return (memcmp(pre, str, lenpre));
}
char *
url_sanitize(char *base_url, char *url, int size)
{
char *newurl;
int base_url_len = strlen(base_url);
if (starts_with(url, "http") == 0) {
newurl = malloc(size+1);
if (newurl == NULL) {
fprintf(stderr, "1 malloc() of %d bytes, failed\n", size);
exit(1);
}
strncpy(newurl, url, size);
newurl[size] = '[=12=]';
} else {
if (starts_with(url, "//") == 0) {
newurl = malloc(size+7);
if (newurl == NULL) {
fprintf(stderr, "2 malloc() of %d bytes, failed\n", size);
exit(1);
}
strncpy(newurl, "https:", 6);
strncpy(newurl+6, url, size);
newurl[size+6] = '[=12=]';
} else {
newurl = malloc(base_url_len + size + 2);
if (newurl == NULL) {
fprintf(stderr, "3 malloc() of %d bytes, failed\n", size);
exit(1);
}
strncpy(newurl, base_url, base_url_len);
strncpy(newurl + base_url_len, url, size);
newurl[size + base_url_len] = '[=12=]';
}
}
return (newurl);
}
char *
html_title_find(char *html)
{
char *newurl, *first, *last;
int size = 0;
first = strstr(html, "<title>");
if (first == NULL)
return (NULL);
first += strlen("<title>");
last = strstr(first, "</title>");
if (last == NULL)
return (NULL);
size = last - first;
newurl = malloc(size+1);
if (newurl == NULL) {
fprintf(stderr, "4 malloc() of %d bytes, failed\n", size);
exit(1);
}
strncpy(newurl, first, size);
newurl[size] = '[=12=]';
return (newurl);
}
void
html_link_find(char *url, char *html)
{
char *first, *last, *newurl;
int size = 0;
first = html;
while (first && last) {
first = strstr(first, "href=\"");
if (first == NULL)
continue;
first += strlen("href=\"");
last = strchr(first, '\"');
if (last == NULL)
continue;
size = last - first;
newurl = url_sanitize(url, first, size);
if (strstr(newurl, "mailto")) {
free(newurl);
continue;
} else {
if (mysql_url_visited_find(newurl) && mysql_url_frontier_find(newurl)) {
mysql_url_frontier_push(newurl);
}
free(newurl);
}
}
}
void
parsed_sites_inc(GlobalInfo *g)
{
g->parsed_sites++;
}
void
html_parse(char *url, char *html)
{
char *title;
title = html_title_find(html);
html_link_find(url, html);
mysql_url_visited_push(url, title);
free(title);
}
/* Die if we get a bad CURLMcode somewhere */
static void
mcode_or_die(const char *where, CURLMcode code)
{
if (CURLM_OK != code) {
const char *s;
switch (code) {
mycase(CURLM_BAD_HANDLE); break;
mycase(CURLM_BAD_EASY_HANDLE); break;
mycase(CURLM_OUT_OF_MEMORY); break;
mycase(CURLM_INTERNAL_ERROR); break;
mycase(CURLM_UNKNOWN_OPTION); break;
mycase(CURLM_LAST); break;
default: s = "CURLM_unknown"; break;
mycase(CURLM_BAD_SOCKET);
fprintf(MSG_OUT, "ERROR: %s returns %s\n", where, s);
/* ignore this error */
return;
}
fprintf(MSG_OUT, "ERROR: %s returns %s\n", where, s);
exit(code);
}
}
void
print_progress(GlobalInfo *g)
{
printf("\rParsed sites: %d, %d parallel connections, %d still running, %d transfers\t",
g->parsed_sites, g->concurrent_connections, g->still_running, g->transfers);
fflush(stdout);
}
void
transfers_inc(GlobalInfo *g)
{
g->transfers++;
print_progress(g);
}
void
transfers_dec(GlobalInfo *g)
{
g->transfers--;
print_progress(g);
}
void
concurrent_connections_inc(GlobalInfo *g)
{
g->concurrent_connections++;
print_progress(g);
}
void
concurrent_connections_dec(GlobalInfo *g)
{
g->concurrent_connections--;
print_progress(g);
}
static void timer_cb(GlobalInfo* g, int revents);
/* Update the timer after curl_multi library does it's thing. Curl will
* inform us through this callback what it wants the new timeout to be,
* after it does some work. */
static int
multi_timer_cb(CURLM *multi, long timeout_ms, GlobalInfo *g)
{
struct itimerspec its;
//fprintf(MSG_OUT, "multi_timer_cb: Setting timeout to %ld ms\n", timeout_ms);
if (timeout_ms > 0) {
its.it_interval.tv_sec = 1;
its.it_interval.tv_nsec = 0;
its.it_value.tv_sec = timeout_ms / 1000;
its.it_value.tv_nsec = (timeout_ms % 1000) * 1000 * 1000;
} else if(timeout_ms == 0) {
/* libcurl wants us to timeout now, however setting both fields of
* new_value.it_value to zero disarms the timer. The closest we can
* do is to schedule the timer to fire in 1 ns. */
its.it_interval.tv_sec = 1;
its.it_interval.tv_nsec = 0;
its.it_value.tv_sec = 0;
its.it_value.tv_nsec = 1;
} else {
memset(&its, 0, sizeof(struct itimerspec));
}
timerfd_settime(g->tfd, /*flags=*/ 0, &its, NULL);
return (0);
}
/* Check for completed transfers, and remove their easy handles */
static void
check_multi_info(GlobalInfo *g)
{
char *eff_url;
CURLMsg *msg;
int msgs_left;
ConnInfo *conn;
CURL *easy;
char *ct;
double time;
double dl;
long header_size;
long response_code;
//CURLcode res;
while ((msg = curl_multi_info_read(g->multi, &msgs_left))) {
if (msg->msg == CURLMSG_DONE) {
easy = msg->easy_handle;
//res = msg->data.result;
curl_easy_getinfo(easy, CURLINFO_PRIVATE, &conn);
curl_easy_getinfo(easy, CURLINFO_EFFECTIVE_URL, &eff_url);
curl_easy_getinfo(easy, CURLINFO_CONTENT_TYPE, &ct);
curl_easy_getinfo(easy, CURLINFO_TOTAL_TIME, &time);
curl_easy_getinfo(easy, CURLINFO_SIZE_DOWNLOAD, &dl);
curl_easy_getinfo(easy, CURLINFO_RESPONSE_CODE, &response_code);
curl_easy_getinfo(easy, CURLINFO_HEADER_SIZE, &header_size);
if (response_code == 200 && dl == 0.0 && (starts_with(ct, "text/html") || starts_with(ct, "text/plain")))
{
/* This should be a response to our HEAD request */
//printf("200 %s header size: %ld download size: %f", eff_url, header_size, dl);
new_body_conn(eff_url, g);
} else if (response_code == 200 && dl > 0.0 && (starts_with(ct, "text/html") || starts_with(ct, "text/plain"))){
/* This should be a response to our GET request */
//printf("%ld %s download size: %f content type: %s\n", response_code, eff_url, dl, ct);
html_parse(eff_url, conn->data);
parsed_sites_inc(g);
}
//fprintf(MSG_OUT, "DONE: %s => (%d) %s\n", eff_url, res, conn->error);
curl_multi_remove_handle(g->multi, easy);
//free(conn->url);
free(conn->data);
curl_easy_cleanup(easy);
transfers_dec(g);
free(conn);
}
}
}
/* Called by libevent when we get action on a multi socket filedescriptor*/
static void
event_cb(GlobalInfo *g, int fd, int revents)
{
CURLMcode rc;
struct itimerspec its;
int action = ((revents & EPOLLIN) ? CURL_CSELECT_IN : 0) |
((revents & EPOLLOUT) ? CURL_CSELECT_OUT : 0);
rc = curl_multi_socket_action(g->multi, fd, action, &g->still_running);
mcode_or_die("event_cb: curl_multi_socket_action", rc);
check_multi_info(g);
if (g->still_running <= 0) {
//fprintf(MSG_OUT, "last transfer done, kill timeout\n");
memset(&its, 0, sizeof(struct itimerspec));
timerfd_settime(g->tfd, 0, &its, NULL);
}
}
/* Called by main loop when our timeout expires */
static void
timer_cb(GlobalInfo* g, int revents)
{
CURLMcode rc;
uint64_t count = 0;
ssize_t err = 0;
err = read(g->tfd, &count, sizeof(uint64_t));
if (err == -1) {
/* Note that we may call the timer callback even if the timerfd isn't
* readable. It's possible that there are multiple events stored in the
* epoll buffer (i.e. the timer may have fired multiple times). The
* event count is cleared after the first call so future events in the
* epoll buffer will fail to read from the timer. */
if (errno == EAGAIN) {
//fprintf(MSG_OUT, "EAGAIN on tfd %d\n", g->tfd);
return;
}
}
if (err != sizeof(uint64_t)) {
fprintf(stderr, "read(tfd) == %ld", err);
perror("read(tfd)");
}
rc = curl_multi_socket_action(g->multi, CURL_SOCKET_TIMEOUT, 0, &g->still_running);
mcode_or_die("timer_cb: curl_multi_socket_action", rc);
check_multi_info(g);
}
/* Assign information to a SockInfo structure */
static void
setsock(SockInfo *f, curl_socket_t s, CURL *e, int act, GlobalInfo *g)
{
struct epoll_event ev;
int kind = ((act & CURL_POLL_IN) ? EPOLLIN : 0) |
((act & CURL_POLL_OUT) ? EPOLLOUT : 0);
if (f->sockfd) {
concurrent_connections_dec(g);
if (epoll_ctl(g->epfd, EPOLL_CTL_DEL, f->sockfd, NULL))
fprintf(stderr, "EPOLL_CTL_DEL failed for fd: %d : %s\n",
f->sockfd, strerror(errno));
}
f->sockfd = s;
f->action = act;
f->easy = e;
ev.events = kind;
ev.data.fd = s;
concurrent_connections_inc(g);
if (epoll_ctl(g->epfd, EPOLL_CTL_ADD, s, &ev)) {
fprintf(stderr, "EPOLL_CTL_ADD failed for fd: %d : %s\n",
s, strerror(errno));
}
}
/* Initialize a new SockInfo structure */
static void
addsock(curl_socket_t s, CURL *easy, int action, GlobalInfo *g)
{
SockInfo *fdp = (SockInfo *)calloc(sizeof(SockInfo), 1);
fdp->global = g;
setsock(fdp, s, easy, action, g);
curl_multi_assign(g->multi, s, fdp);
}
static size_t
write_cb(void *contents, size_t size, size_t nmemb, void *p)
{
ConnInfo *conn = (ConnInfo *)p;
size_t realsize = size * nmemb;
conn->data = realloc(conn->data, conn->size + realsize + 1);
if (conn->data == NULL) {
/* out of memory! */
printf("not enough memory (realloc returned NULL)\n");
return 0;
}
memcpy(&(conn->data[conn->size]), contents, realsize);
conn->size += realsize;
conn->data[conn->size] = 0;
return realsize;
}
/* Create a new easy handle, and add it to the global curl_multi */
int
new_head_conn(char *url, GlobalInfo *g)
{
ConnInfo *conn;
CURLMcode rc;
conn = (ConnInfo*)calloc(1, sizeof(ConnInfo));
conn->error[0]='[=12=]';
conn->global = g;
conn->easy = curl_easy_init();
if (!conn->easy) {
fprintf(MSG_OUT, "curl_easy_init() failed, exiting!\n");
exit(2);
}
transfers_inc(g);
conn->global = g;
conn->url = url;
curl_easy_setopt(conn->easy, CURLOPT_URL, conn->url);
curl_easy_setopt(conn->easy, CURLOPT_WRITEFUNCTION, write_cb);
curl_easy_setopt(conn->easy, CURLOPT_WRITEDATA, conn);
curl_easy_setopt(conn->easy, CURLOPT_FOLLOWLOCATION, 1L);
curl_easy_setopt(conn->easy, CURLOPT_ERRORBUFFER, conn->error);
curl_easy_setopt(conn->easy, CURLOPT_PRIVATE, conn);
curl_easy_setopt(conn->easy, CURLOPT_NOPROGRESS, 1L);
curl_easy_setopt(conn->easy, CURLOPT_PROGRESSDATA, conn);
curl_easy_setopt(conn->easy, CURLOPT_FOLLOWLOCATION, 1L);
curl_easy_setopt(conn->easy, CURLOPT_LOW_SPEED_TIME, 3L);
curl_easy_setopt(conn->easy, CURLOPT_LOW_SPEED_LIMIT, 100L);
curl_easy_setopt(conn->easy, CURLOPT_CONNECTTIMEOUT, 10L);
curl_easy_setopt(conn->easy, CURLOPT_CLOSESOCKETDATA, g);
curl_easy_setopt(conn->easy, CURLOPT_NOBODY, 1L);
rc = curl_multi_add_handle(g->multi, conn->easy);
mcode_or_die("new_conn: curl_multi_add_handle", rc);
/* note that the add_handle() will set a time-out to trigger very soon so
that the necessary socket_action() call will be called by this app */
return (0);
}
/* Create a new easy handle, and add it to the global curl_multi */
int
new_body_conn(char *url, GlobalInfo *g)
{
ConnInfo *conn;
CURLMcode rc;
conn = (ConnInfo*)calloc(1, sizeof(ConnInfo));
conn->error[0]='[=12=]';
conn->global = g;
conn->easy = curl_easy_init();
if (!conn->easy) {
fprintf(MSG_OUT, "curl_easy_init() failed, exiting!\n");
exit(2);
}
transfers_inc(g);
conn->global = g;
conn->url = url;
curl_easy_setopt(conn->easy, CURLOPT_URL, conn->url);
curl_easy_setopt(conn->easy, CURLOPT_WRITEFUNCTION, write_cb);
curl_easy_setopt(conn->easy, CURLOPT_WRITEDATA, conn);
curl_easy_setopt(conn->easy, CURLOPT_FOLLOWLOCATION, 1L);
curl_easy_setopt(conn->easy, CURLOPT_ERRORBUFFER, conn->error);
curl_easy_setopt(conn->easy, CURLOPT_PRIVATE, conn);
curl_easy_setopt(conn->easy, CURLOPT_NOPROGRESS, 1L);
curl_easy_setopt(conn->easy, CURLOPT_PROGRESSDATA, conn);
curl_easy_setopt(conn->easy, CURLOPT_FOLLOWLOCATION, 1L);
curl_easy_setopt(conn->easy, CURLOPT_LOW_SPEED_TIME, 3L);
curl_easy_setopt(conn->easy, CURLOPT_LOW_SPEED_LIMIT, 100L);
curl_easy_setopt(conn->easy, CURLOPT_CONNECTTIMEOUT, 10L);
curl_easy_setopt(conn->easy, CURLOPT_CLOSESOCKETDATA, g);
rc = curl_multi_add_handle(g->multi, conn->easy);
mcode_or_die("new_conn: curl_multi_add_handle", rc);
/* note that the add_handle() will set a time-out to trigger very soon so
that the necessary socket_action() call will be called by this app */
return (0);
}
/* Clean up the SockInfo structure */
static void
remsock(SockInfo *f, GlobalInfo* g)
{
if (f) {
if (f->sockfd) {
concurrent_connections_dec(g);
if (epoll_ctl(g->epfd, EPOLL_CTL_DEL, f->sockfd, NULL))
fprintf(stderr, "EPOLL_CTL_DEL failed for fd: %d : %s\n",
f->sockfd, strerror(errno));
}
free(f);
}
}
/* CURLMOPT_SOCKETFUNCTION */
static int
sock_cb(CURL *e, curl_socket_t s, int what, void *cbp, void *sockp)
{
GlobalInfo *g = (GlobalInfo*) cbp;
SockInfo *fdp = (SockInfo*) sockp;
if (what == CURL_POLL_REMOVE) {
remsock(fdp, g);
} else {
if (g->concurrent_connections < MAX_CONNECTIONS){
if (!fdp) {
addsock(s, e, what, g);
} else {
setsock(fdp, s, e, what, g);
}
}
}
return (0);
}
/* CURLMOPT_SOCKETFUNCTION */
static int
end_sock_cb(CURL *e, curl_socket_t s, int what, void *cbp, void *sockp)
{
GlobalInfo *g = (GlobalInfo*) cbp;
SockInfo *fdp = (SockInfo*) sockp;
if (what == CURL_POLL_REMOVE) {
remsock(fdp, g);
}
return (0);
}
int should_exit = 0;
void
signal_handler(int signo)
{
should_exit = 1;
}
void *
crawler_init()
{
GlobalInfo g;
struct itimerspec its;
struct epoll_event ev;
struct epoll_event events[10000];
memset(&g, 0, sizeof(GlobalInfo));
g.transfers = 0;
g.parsed_sites = 0;
g.epfd = epoll_create1(EPOLL_CLOEXEC);
if (g.epfd == -1) {
perror("epoll_create1 failed\n");
exit(1);
}
g.tfd = timerfd_create(CLOCK_MONOTONIC, TFD_NONBLOCK | TFD_CLOEXEC);
if (g.tfd == -1) {
perror("timerfd_create failed\n");
exit(1);
}
memset(&its, 0, sizeof(struct itimerspec));
its.it_interval.tv_sec = 1;
its.it_value.tv_sec = 1;
timerfd_settime(g.tfd, 0, &its, NULL);
ev.events = EPOLLIN;
ev.data.fd = g.tfd;
epoll_ctl(g.epfd, EPOLL_CTL_ADD, g.tfd, &ev);
curl_global_init(CURL_GLOBAL_DEFAULT);
g.multi = curl_multi_init();
/* setup the generic multi interface options we want */
curl_multi_setopt(g.multi, CURLMOPT_SOCKETFUNCTION, sock_cb);
curl_multi_setopt(g.multi, CURLMOPT_SOCKETDATA, &g);
curl_multi_setopt(g.multi, CURLMOPT_TIMERFUNCTION, multi_timer_cb);
curl_multi_setopt(g.multi, CURLMOPT_TIMERDATA, &g);
/* we don't call any curl_multi_socket*() function yet as we have no handles added! */
//printf("Starting crawler...\n");
while (!should_exit) {
int idx;
int err = epoll_wait(g.epfd, events, sizeof(events)/sizeof(struct epoll_event), 10000);
char *url;
url = mysql_url_frontier_pop();
new_head_conn(url, &g);
if (err == -1) {
if (errno == EINTR) {
fprintf(MSG_OUT, "note: wait interrupted\n");
continue;
} else {
perror("epoll_wait");
exit(1);
}
}
for (idx = 0; idx < err; ++idx) {
if (events[idx].data.fd == g.tfd) {
timer_cb(&g, events[idx].events);
} else {
event_cb(&g, events[idx].data.fd, events[idx].events);
}
}
}
fprintf(MSG_OUT, "Exiting normally.\n");
fflush(MSG_OUT);
curl_multi_setopt(g.multi, CURLMOPT_SOCKETFUNCTION, end_sock_cb);
while (g.concurrent_connections > 0 || g.transfers > 0)
{
int idx;
int err = epoll_wait(g.epfd, events, sizeof(events)/sizeof(struct epoll_event), 10000);
if (err == -1) {
if (errno == EINTR) {
fprintf(MSG_OUT, "note: wait interrupted\n");
continue;
} else {
perror("epoll_wait");
exit(1);
}
}
for (idx = 0; idx < err; ++idx) {
if (events[idx].data.fd == g.tfd) {
timer_cb(&g, events[idx].events);
} else {
event_cb(&g, events[idx].data.fd, events[idx].events);
}
}
}
fprintf(MSG_OUT, "Finished all in progress downloads.\n");
fflush(MSG_OUT);
curl_multi_cleanup(g.multi);
curl_global_cleanup();
return (NULL);
}
int
main(int argc, char **argv)
{
int cleanup = 0, opt, ret;
should_exit = 0;
signal(SIGINT, signal_handler);
signal(SIGKILL, signal_handler);
mysql_start();
crawler_init();
mysql_stop();
printf("Exiting.\n");
return (0);
}
典型输出为：
Parsed sites: 42, 6 parallel connections, 4 still running, 6 transfersssExiting normally.
Parsed sites: 48, 0 parallel connections, 0 still running, 0 transfers Finished all in progress downloads.
Exiting.
如您所见，该应用并未达到硬编码的最大并行连接数。差远了。而且它只消耗大约 3 Mbps 的带宽。
为什么？我不明白为什么仅添加 MySQL 代码会导致程序无法 运行 发挥其全部潜力。无法达到硬编码的最大并行连接数。
知道这段代码有什么问题吗？
编辑
这是一些 gprof 输出：
Flat profile:
Each sample counts as 0.01 seconds.
% cumulative self self total
time seconds seconds calls ms/call ms/call name
44.49 0.04 0.04 12668 0.00 0.00 mysql_url_visited_find
33.37 0.07 0.03 12668 0.00 0.00 mysql_url_frontier_push
11.12 0.08 0.01 12668 0.00 0.00 mysql_url_frontier_find
11.12 0.09 0.01 100 0.10 0.90 html_link_find
0.00 0.09 0.00 17355 0.00 0.00 starts_with
0.00 0.09 0.00 12669 0.00 0.00 url_sanitize
0.00 0.09 0.00 2651 0.00 0.00 mcode_or_die
0.00 0.09 0.00 2432 0.00 0.04 check_multi_info
0.00 0.09 0.00 2420 0.00 0.04 event_cb
0.00 0.09 0.00 1288 0.00 0.00 print_progress
0.00 0.09 0.00 425 0.00 0.00 concurrent_connections_dec
0.00 0.09 0.00 425 0.00 0.00 concurrent_connections_inc
0.00 0.09 0.00 425 0.00 0.00 setsock
0.00 0.09 0.00 303 0.00 0.00 remsock
0.00 0.09 0.00 299 0.00 0.00 addsock
0.00 0.09 0.00 219 0.00 0.00 transfers_dec
0.00 0.09 0.00 219 0.00 0.00 transfers_inc
0.00 0.09 0.00 116 0.00 0.00 mysql_url_frontier_pop
0.00 0.09 0.00 116 0.00 0.00 new_head_conn
0.00 0.09 0.00 103 0.00 0.00 new_body_conn
0.00 0.09 0.00 100 0.00 0.90 html_parse
0.00 0.09 0.00 100 0.00 0.00 html_title_find
0.00 0.09 0.00 100 0.00 0.00 mysql_url_visited_push
0.00 0.09 0.00 100 0.00 0.00 parsed_sites_inc
0.00 0.09 0.00 32 0.00 0.01 timer_cb
0.00 0.09 0.00 1 0.00 90.09 crawler_init
0.00 0.09 0.00 1 0.00 0.00 mysql_start
0.00 0.09 0.00 1 0.00 0.00 mysql_stop
您的所有 mysql 代码都是同步的 — 在您等待 mysql 响应查询或确认插入时，您的事件循环没有做任何事情，并且您花费了 90您在 mysql 相关功能中花费的时间百分比。您没有接近并发连接限制，因为数据库瓶颈限制了您发出新 HTTP 请求的速率，并且在您等待D B。
解决这个问题并不简单，但有一些一般性建议：
尽可能避免与数据库对话。 'frontier' 内容看起来像是可以保持在进程中的状态。
使用异步库与数据库交互并将其集成到您的事件循环中，这样您就可以在数据库查询未完成的情况下继续其他工作。
确保你的数据库有必要的索引，或者考虑使用像 redis 这样的东西而不是 mysql，其中访问模式更明确，简单的访问非常快，你可以'真的 "accidentally" 写了一个性能很差的查询。
我有一个使用 libcurl 和 epoll 来抓取网站的工作代码：
https://github.com/JamesRead5737/libcurlmemoryleak/blob/master/crawler.c
典型输出为：
Parsed sites: 0, 1024 parallel connections, 10989 still running, 10989 transfers Exiting normally.
Parsed sites: 0, 0 parallel connections, 0 still running, 0 transfersrsrFinished all in progress downloads.
Exiting.
如您所见，代码达到了 1024 个并行连接的硬编码限制。这在我的专用服务器上最多消耗 3 Gbps。
像这样添加 Mysql 代码：
#include <errno.h>
#include <fcntl.h>
#include <signal.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/epoll.h>
#include <sys/stat.h>
#include <sys/time.h>
#include <sys/timerfd.h>
#include <sys/types.h>
#include <sys/resource.h>
#include <sys/socket.h>
#include <arpa/inet.h>
#include <time.h>
#include <unistd.h>
#include <pthread.h>
#include <netdb.h>
#include <time.h>
#include <netinet/in.h>
#include <mysql.h>
#include <curl/curl.h>
#define MSG_OUT stdout
#define DEFAULT_QUEUE_LENGTH 10000
#define mycase(code) \
case code: s = __STRING(code)
#define MAX_CONNECTIONS 1024
MYSQL *mysql_con;
/* Global information, common to all connections */
typedef struct _GlobalInfo
{
int epfd; /* epoll filedescriptor */
int tfd; /* timer filedescriptor */
CURLM *multi;
int still_running;
pthread_mutex_t lock;
int concurrent_connections;
pthread_mutex_t parsed_lock;
int parsed_sites;
int transfers;
} GlobalInfo;
int new_body_conn(char *url, GlobalInfo *g);
/* Information associated with a specific easy handle */
typedef struct _ConnInfo
{
CURL *easy;
char *url;
GlobalInfo *global;
char error[CURL_ERROR_SIZE];
size_t size;
char *data;
} ConnInfo;
/* Information associated with a specific socket */
typedef struct _SockInfo
{
curl_socket_t sockfd;
CURL *easy;
int action;
long timeout;
GlobalInfo *global;
} SockInfo;
void
mysql_stop()
{
mysql_close(mysql_con);
}
void
mysql_start()
{
mysql_con = mysql_init(NULL);
if (mysql_con == NULL)
{
fprintf(stderr, "%s\n", mysql_error(mysql_con));
exit(1);
}
if (mysql_real_connect(mysql_con, "localhost", "crawler", "password", "crawl", 0, NULL, 0) == NULL)
{
fprintf(stderr, "%s\n", mysql_error(mysql_con));
exit(EXIT_FAILURE);
}
if (mysql_query(mysql_con, "CREATE TABLE IF NOT EXISTS `frontier` (`url` varchar(2084) NOT NULL, `id` int NOT NULL AUTO_INCREMENT, PRIMARY KEY (`id`), UNIQUE KEY `url` (`url`), KEY `url_2` (`url`)) ENGINE=InnoDB AUTO_INCREMENT=1"))
{
fprintf(stderr, "%s\n", mysql_error(mysql_con));
mysql_stop();
exit(1);
}
if (mysql_query(mysql_con, "CREATE TABLE IF NOT EXISTS `crawled` (`id` int NOT NULL AUTO_INCREMENT, `url` varchar(2084) DEFAULT NULL, `title` varchar(768) DEFAULT NULL, `date` varchar(128) DEFAULT NULL, `last_modified` varchar(128) DEFAULT NULL, `links` int DEFAULT NULL, `backlinks` int DEFAULT NULL, `http_code` int DEFAULT NULL, PRIMARY KEY (`id`), UNIQUE KEY `url` (`url`), KEY `http_code` (`http_code`), KEY `title` (`title`)) ENGINE=InnoDB AUTO_INCREMENT=1"))
{
fprintf(stderr, "%s\n", mysql_error(mysql_con));
mysql_stop();
exit(1);
}
if (mysql_query(mysql_con, "CREATE TABLE IF NOT EXISTS `emails` (`email` varchar(2084) NOT NULL, `id` int NOT NULL AUTO_INCREMENT, PRIMARY KEY (`id`), UNIQUE KEY `email` (`email`), KEY `email_2` (`email`)) ENGINE=InnoDB AUTO_INCREMENT=737 DEFAULT CHARSET=latin1"))
{
fprintf(stderr, "%s\n", mysql_error(mysql_con));
mysql_stop();
exit(1);
}
}
void
mysql_url_visited_push(char *url, char *title)
{
char sql[8192];
char escaped_url[(strlen(url)*2)+1];
if (!mysql_real_escape_string(mysql_con, escaped_url, url, strlen(url)))
{
}
if (title != NULL)
{
char escaped_title[(strlen(title)*2)+1];
if (!mysql_real_escape_string(mysql_con, escaped_title, title, strlen(title)))
{
}
sprintf(sql, "INSERT INTO crawled (url, title) VALUES ('%s', '%s')", escaped_url, escaped_title);
} else {
sprintf(sql, "INSERT INTO crawled (url, title) VALUES ('%s', '%s')", escaped_url, title);
}
if (mysql_query(mysql_con, sql))
{
fprintf(stderr, "mysql_url_visited_push sql=%s %s\n", sql, mysql_error(mysql_con));
}
}
int
mysql_url_visited_find(char *url)
{
char sql[8192];
char escaped_url[(strlen(url)*2)+1];
if (!mysql_real_escape_string(mysql_con, escaped_url, url, strlen(url)))
{
}
sprintf(sql, "SELECT * FROM crawled WHERE url = '%s'", escaped_url);
if (mysql_query(mysql_con, sql))
{
fprintf(stderr, "mysql_url_visited_find sql=%s %s\n", sql, mysql_error(mysql_con));
} else {
MYSQL_RES *result = mysql_store_result(mysql_con);
if (!result)
{
fprintf(stderr, "%s\n", mysql_error(mysql_con));
} else if (result > 0){
mysql_free_result(result);
return 1;
} else if (result == 0) {
mysql_free_result(result);
return 0;
}
}
}
int
mysql_url_frontier_find(char *url)
{
char sql[8192];
char escaped_url[(strlen(url)*2)+1];
if (!mysql_real_escape_string(mysql_con, escaped_url, url, strlen(url)))
{
}
sprintf(sql, "SELECT * FROM frontier WHERE url = '%s'", escaped_url);
if (mysql_query(mysql_con, sql))
{
fprintf(stderr, "mysql_url_frontier_find sql=%s %s\n", sql, mysql_error(mysql_con));
} else {
MYSQL_RES *result = mysql_store_result(mysql_con);
if (!result)
{
fprintf(stderr, "%s\n", mysql_error(mysql_con));
} else if (result > 0){
mysql_free_result(result);
return 1;
} else if (result == 0) {
mysql_free_result(result);
return 0;
}
}
}
void
mysql_url_frontier_push(char *url)
{
char sql[8192];
char escaped_url[(strlen(url)*2)+1];
if (!mysql_real_escape_string(mysql_con, escaped_url, url, strlen(url)))
{
}
sprintf(sql, "INSERT IGNORE INTO frontier (url) VALUES ('%s')", escaped_url);
if (mysql_query(mysql_con, sql))
{
fprintf(stderr, "mysql_url_frontier_push sql=%s %s\n", sql, mysql_error(mysql_con));
}
}
char *
mysql_url_frontier_pop()
{
char *url = NULL;
char sql[8192];
if (mysql_query(mysql_con, "SELECT url FROM frontier ORDER BY id") == 0)
{
MYSQL_ROW row;
MYSQL_RES *result = mysql_store_result(mysql_con);
if (result == NULL)
{
fprintf(stderr, "mysql_url_frontier_pop mysql_store_result sql=%s %s\n", sql, mysql_error(mysql_con));
exit(EXIT_FAILURE);
}
if ((row = mysql_fetch_row(result)))
{
url = strdup(row[0]);
char escaped_url[(strlen(url)*2)+1];
if (!mysql_real_escape_string(mysql_con, escaped_url, url, strlen(url)))
{
}
sprintf( sql, "DELETE FROM frontier WHERE url = '%s'", escaped_url);
if (mysql_query(mysql_con, sql))
{
fprintf(stderr, "mysql_url_frontier_pop mysql_query sql=%s %s\n", sql, mysql_error(mysql_con));
exit(EXIT_FAILURE);
}
}
mysql_free_result(result);
}
return url;
}
int
starts_with(const char *str, const char *pre)
{
size_t lenstr;
size_t lenpre;
if (str == NULL || pre == NULL)
return (-1);
lenstr = strlen(str);
lenpre = strlen(pre);
if (lenstr < lenpre)
return (-1);
return (memcmp(pre, str, lenpre));
}
char *
url_sanitize(char *base_url, char *url, int size)
{
char *newurl;
int base_url_len = strlen(base_url);
if (starts_with(url, "http") == 0) {
newurl = malloc(size+1);
if (newurl == NULL) {
fprintf(stderr, "1 malloc() of %d bytes, failed\n", size);
exit(1);
}
strncpy(newurl, url, size);
newurl[size] = '[=12=]';
} else {
if (starts_with(url, "//") == 0) {
newurl = malloc(size+7);
if (newurl == NULL) {
fprintf(stderr, "2 malloc() of %d bytes, failed\n", size);
exit(1);
}
strncpy(newurl, "https:", 6);
strncpy(newurl+6, url, size);
newurl[size+6] = '[=12=]';
} else {
newurl = malloc(base_url_len + size + 2);
if (newurl == NULL) {
fprintf(stderr, "3 malloc() of %d bytes, failed\n", size);
exit(1);
}
strncpy(newurl, base_url, base_url_len);
strncpy(newurl + base_url_len, url, size);
newurl[size + base_url_len] = '[=12=]';
}
}
return (newurl);
}
char *
html_title_find(char *html)
{
char *newurl, *first, *last;
int size = 0;
first = strstr(html, "<title>");
if (first == NULL)
return (NULL);
first += strlen("<title>");
last = strstr(first, "</title>");
if (last == NULL)
return (NULL);
size = last - first;
newurl = malloc(size+1);
if (newurl == NULL) {
fprintf(stderr, "4 malloc() of %d bytes, failed\n", size);
exit(1);
}
strncpy(newurl, first, size);
newurl[size] = '[=12=]';
return (newurl);
}
void
html_link_find(char *url, char *html)
{
char *first, *last, *newurl;
int size = 0;
first = html;
while (first && last) {
first = strstr(first, "href=\"");
if (first == NULL)
continue;
first += strlen("href=\"");
last = strchr(first, '\"');
if (last == NULL)
continue;
size = last - first;
newurl = url_sanitize(url, first, size);
if (strstr(newurl, "mailto")) {
free(newurl);
continue;
} else {
if (mysql_url_visited_find(newurl) && mysql_url_frontier_find(newurl)) {
mysql_url_frontier_push(newurl);
}
free(newurl);
}
}
}
void
parsed_sites_inc(GlobalInfo *g)
{
g->parsed_sites++;
}
void
html_parse(char *url, char *html)
{
char *title;
title = html_title_find(html);
html_link_find(url, html);
mysql_url_visited_push(url, title);
free(title);
}
/* Die if we get a bad CURLMcode somewhere */
static void
mcode_or_die(const char *where, CURLMcode code)
{
if (CURLM_OK != code) {
const char *s;
switch (code) {
mycase(CURLM_BAD_HANDLE); break;
mycase(CURLM_BAD_EASY_HANDLE); break;
mycase(CURLM_OUT_OF_MEMORY); break;
mycase(CURLM_INTERNAL_ERROR); break;
mycase(CURLM_UNKNOWN_OPTION); break;
mycase(CURLM_LAST); break;
default: s = "CURLM_unknown"; break;
mycase(CURLM_BAD_SOCKET);
fprintf(MSG_OUT, "ERROR: %s returns %s\n", where, s);
/* ignore this error */
return;
}
fprintf(MSG_OUT, "ERROR: %s returns %s\n", where, s);
exit(code);
}
}
void
print_progress(GlobalInfo *g)
{
printf("\rParsed sites: %d, %d parallel connections, %d still running, %d transfers\t",
g->parsed_sites, g->concurrent_connections, g->still_running, g->transfers);
fflush(stdout);
}
void
transfers_inc(GlobalInfo *g)
{
g->transfers++;
print_progress(g);
}
void
transfers_dec(GlobalInfo *g)
{
g->transfers--;
print_progress(g);
}
void
concurrent_connections_inc(GlobalInfo *g)
{
g->concurrent_connections++;
print_progress(g);
}
void
concurrent_connections_dec(GlobalInfo *g)
{
g->concurrent_connections--;
print_progress(g);
}
static void timer_cb(GlobalInfo* g, int revents);
/* Update the timer after curl_multi library does it's thing. Curl will
* inform us through this callback what it wants the new timeout to be,
* after it does some work. */
static int
multi_timer_cb(CURLM *multi, long timeout_ms, GlobalInfo *g)
{
struct itimerspec its;
//fprintf(MSG_OUT, "multi_timer_cb: Setting timeout to %ld ms\n", timeout_ms);
if (timeout_ms > 0) {
its.it_interval.tv_sec = 1;
its.it_interval.tv_nsec = 0;
its.it_value.tv_sec = timeout_ms / 1000;
its.it_value.tv_nsec = (timeout_ms % 1000) * 1000 * 1000;
} else if(timeout_ms == 0) {
/* libcurl wants us to timeout now, however setting both fields of
* new_value.it_value to zero disarms the timer. The closest we can
* do is to schedule the timer to fire in 1 ns. */
its.it_interval.tv_sec = 1;
its.it_interval.tv_nsec = 0;
its.it_value.tv_sec = 0;
its.it_value.tv_nsec = 1;
} else {
memset(&its, 0, sizeof(struct itimerspec));
}
timerfd_settime(g->tfd, /*flags=*/ 0, &its, NULL);
return (0);
}
/* Check for completed transfers, and remove their easy handles */
static void
check_multi_info(GlobalInfo *g)
{
char *eff_url;
CURLMsg *msg;
int msgs_left;
ConnInfo *conn;
CURL *easy;
char *ct;
double time;
double dl;
long header_size;
long response_code;
//CURLcode res;
while ((msg = curl_multi_info_read(g->multi, &msgs_left))) {
if (msg->msg == CURLMSG_DONE) {
easy = msg->easy_handle;
//res = msg->data.result;
curl_easy_getinfo(easy, CURLINFO_PRIVATE, &conn);
curl_easy_getinfo(easy, CURLINFO_EFFECTIVE_URL, &eff_url);
curl_easy_getinfo(easy, CURLINFO_CONTENT_TYPE, &ct);
curl_easy_getinfo(easy, CURLINFO_TOTAL_TIME, &time);
curl_easy_getinfo(easy, CURLINFO_SIZE_DOWNLOAD, &dl);
curl_easy_getinfo(easy, CURLINFO_RESPONSE_CODE, &response_code);
curl_easy_getinfo(easy, CURLINFO_HEADER_SIZE, &header_size);
if (response_code == 200 && dl == 0.0 && (starts_with(ct, "text/html") || starts_with(ct, "text/plain")))
{
/* This should be a response to our HEAD request */
//printf("200 %s header size: %ld download size: %f", eff_url, header_size, dl);
new_body_conn(eff_url, g);
} else if (response_code == 200 && dl > 0.0 && (starts_with(ct, "text/html") || starts_with(ct, "text/plain"))){
/* This should be a response to our GET request */
//printf("%ld %s download size: %f content type: %s\n", response_code, eff_url, dl, ct);
html_parse(eff_url, conn->data);
parsed_sites_inc(g);
}
//fprintf(MSG_OUT, "DONE: %s => (%d) %s\n", eff_url, res, conn->error);
curl_multi_remove_handle(g->multi, easy);
//free(conn->url);
free(conn->data);
curl_easy_cleanup(easy);
transfers_dec(g);
free(conn);
}
}
}
/* Called by libevent when we get action on a multi socket filedescriptor*/
static void
event_cb(GlobalInfo *g, int fd, int revents)
{
CURLMcode rc;
struct itimerspec its;
int action = ((revents & EPOLLIN) ? CURL_CSELECT_IN : 0) |
((revents & EPOLLOUT) ? CURL_CSELECT_OUT : 0);
rc = curl_multi_socket_action(g->multi, fd, action, &g->still_running);
mcode_or_die("event_cb: curl_multi_socket_action", rc);
check_multi_info(g);
if (g->still_running <= 0) {
//fprintf(MSG_OUT, "last transfer done, kill timeout\n");
memset(&its, 0, sizeof(struct itimerspec));
timerfd_settime(g->tfd, 0, &its, NULL);
}
}
/* Called by main loop when our timeout expires */
static void
timer_cb(GlobalInfo* g, int revents)
{
CURLMcode rc;
uint64_t count = 0;
ssize_t err = 0;
err = read(g->tfd, &count, sizeof(uint64_t));
if (err == -1) {
/* Note that we may call the timer callback even if the timerfd isn't
* readable. It's possible that there are multiple events stored in the
* epoll buffer (i.e. the timer may have fired multiple times). The
* event count is cleared after the first call so future events in the
* epoll buffer will fail to read from the timer. */
if (errno == EAGAIN) {
//fprintf(MSG_OUT, "EAGAIN on tfd %d\n", g->tfd);
return;
}
}
if (err != sizeof(uint64_t)) {
fprintf(stderr, "read(tfd) == %ld", err);
perror("read(tfd)");
}
rc = curl_multi_socket_action(g->multi, CURL_SOCKET_TIMEOUT, 0, &g->still_running);
mcode_or_die("timer_cb: curl_multi_socket_action", rc);
check_multi_info(g);
}
/* Assign information to a SockInfo structure */
static void
setsock(SockInfo *f, curl_socket_t s, CURL *e, int act, GlobalInfo *g)
{
struct epoll_event ev;
int kind = ((act & CURL_POLL_IN) ? EPOLLIN : 0) |
((act & CURL_POLL_OUT) ? EPOLLOUT : 0);
if (f->sockfd) {
concurrent_connections_dec(g);
if (epoll_ctl(g->epfd, EPOLL_CTL_DEL, f->sockfd, NULL))
fprintf(stderr, "EPOLL_CTL_DEL failed for fd: %d : %s\n",
f->sockfd, strerror(errno));
}
f->sockfd = s;
f->action = act;
f->easy = e;
ev.events = kind;
ev.data.fd = s;
concurrent_connections_inc(g);
if (epoll_ctl(g->epfd, EPOLL_CTL_ADD, s, &ev)) {
fprintf(stderr, "EPOLL_CTL_ADD failed for fd: %d : %s\n",
s, strerror(errno));
}
}
/* Initialize a new SockInfo structure */
static void
addsock(curl_socket_t s, CURL *easy, int action, GlobalInfo *g)
{
SockInfo *fdp = (SockInfo *)calloc(sizeof(SockInfo), 1);
fdp->global = g;
setsock(fdp, s, easy, action, g);
curl_multi_assign(g->multi, s, fdp);
}
static size_t
write_cb(void *contents, size_t size, size_t nmemb, void *p)
{
ConnInfo *conn = (ConnInfo *)p;
size_t realsize = size * nmemb;
conn->data = realloc(conn->data, conn->size + realsize + 1);
if (conn->data == NULL) {
/* out of memory! */
printf("not enough memory (realloc returned NULL)\n");
return 0;
}
memcpy(&(conn->data[conn->size]), contents, realsize);
conn->size += realsize;
conn->data[conn->size] = 0;
return realsize;
}
/* Create a new easy handle, and add it to the global curl_multi */
int
new_head_conn(char *url, GlobalInfo *g)
{
ConnInfo *conn;
CURLMcode rc;
conn = (ConnInfo*)calloc(1, sizeof(ConnInfo));
conn->error[0]='[=12=]';
conn->global = g;
conn->easy = curl_easy_init();
if (!conn->easy) {
fprintf(MSG_OUT, "curl_easy_init() failed, exiting!\n");
exit(2);
}
transfers_inc(g);
conn->global = g;
conn->url = url;
curl_easy_setopt(conn->easy, CURLOPT_URL, conn->url);
curl_easy_setopt(conn->easy, CURLOPT_WRITEFUNCTION, write_cb);
curl_easy_setopt(conn->easy, CURLOPT_WRITEDATA, conn);
curl_easy_setopt(conn->easy, CURLOPT_FOLLOWLOCATION, 1L);
curl_easy_setopt(conn->easy, CURLOPT_ERRORBUFFER, conn->error);
curl_easy_setopt(conn->easy, CURLOPT_PRIVATE, conn);
curl_easy_setopt(conn->easy, CURLOPT_NOPROGRESS, 1L);
curl_easy_setopt(conn->easy, CURLOPT_PROGRESSDATA, conn);
curl_easy_setopt(conn->easy, CURLOPT_FOLLOWLOCATION, 1L);
curl_easy_setopt(conn->easy, CURLOPT_LOW_SPEED_TIME, 3L);
curl_easy_setopt(conn->easy, CURLOPT_LOW_SPEED_LIMIT, 100L);
curl_easy_setopt(conn->easy, CURLOPT_CONNECTTIMEOUT, 10L);
curl_easy_setopt(conn->easy, CURLOPT_CLOSESOCKETDATA, g);
curl_easy_setopt(conn->easy, CURLOPT_NOBODY, 1L);
rc = curl_multi_add_handle(g->multi, conn->easy);
mcode_or_die("new_conn: curl_multi_add_handle", rc);
/* note that the add_handle() will set a time-out to trigger very soon so
that the necessary socket_action() call will be called by this app */
return (0);
}
/* Create a new easy handle, and add it to the global curl_multi */
int
new_body_conn(char *url, GlobalInfo *g)
{
ConnInfo *conn;
CURLMcode rc;
conn = (ConnInfo*)calloc(1, sizeof(ConnInfo));
conn->error[0]='[=12=]';
conn->global = g;
conn->easy = curl_easy_init();
if (!conn->easy) {
fprintf(MSG_OUT, "curl_easy_init() failed, exiting!\n");
exit(2);
}
transfers_inc(g);
conn->global = g;
conn->url = url;
curl_easy_setopt(conn->easy, CURLOPT_URL, conn->url);
curl_easy_setopt(conn->easy, CURLOPT_WRITEFUNCTION, write_cb);
curl_easy_setopt(conn->easy, CURLOPT_WRITEDATA, conn);
curl_easy_setopt(conn->easy, CURLOPT_FOLLOWLOCATION, 1L);
curl_easy_setopt(conn->easy, CURLOPT_ERRORBUFFER, conn->error);
curl_easy_setopt(conn->easy, CURLOPT_PRIVATE, conn);
curl_easy_setopt(conn->easy, CURLOPT_NOPROGRESS, 1L);
curl_easy_setopt(conn->easy, CURLOPT_PROGRESSDATA, conn);
curl_easy_setopt(conn->easy, CURLOPT_FOLLOWLOCATION, 1L);
curl_easy_setopt(conn->easy, CURLOPT_LOW_SPEED_TIME, 3L);
curl_easy_setopt(conn->easy, CURLOPT_LOW_SPEED_LIMIT, 100L);
curl_easy_setopt(conn->easy, CURLOPT_CONNECTTIMEOUT, 10L);
curl_easy_setopt(conn->easy, CURLOPT_CLOSESOCKETDATA, g);
rc = curl_multi_add_handle(g->multi, conn->easy);
mcode_or_die("new_conn: curl_multi_add_handle", rc);
/* note that the add_handle() will set a time-out to trigger very soon so
that the necessary socket_action() call will be called by this app */
return (0);
}
/* Clean up the SockInfo structure */
static void
remsock(SockInfo *f, GlobalInfo* g)
{
if (f) {
if (f->sockfd) {
concurrent_connections_dec(g);
if (epoll_ctl(g->epfd, EPOLL_CTL_DEL, f->sockfd, NULL))
fprintf(stderr, "EPOLL_CTL_DEL failed for fd: %d : %s\n",
f->sockfd, strerror(errno));
}
free(f);
}
}
/* CURLMOPT_SOCKETFUNCTION */
static int
sock_cb(CURL *e, curl_socket_t s, int what, void *cbp, void *sockp)
{
GlobalInfo *g = (GlobalInfo*) cbp;
SockInfo *fdp = (SockInfo*) sockp;
if (what == CURL_POLL_REMOVE) {
remsock(fdp, g);
} else {
if (g->concurrent_connections < MAX_CONNECTIONS){
if (!fdp) {
addsock(s, e, what, g);
} else {
setsock(fdp, s, e, what, g);
}
}
}
return (0);
}
/* CURLMOPT_SOCKETFUNCTION */
static int
end_sock_cb(CURL *e, curl_socket_t s, int what, void *cbp, void *sockp)
{
GlobalInfo *g = (GlobalInfo*) cbp;
SockInfo *fdp = (SockInfo*) sockp;
if (what == CURL_POLL_REMOVE) {
remsock(fdp, g);
}
return (0);
}
int should_exit = 0;
void
signal_handler(int signo)
{
should_exit = 1;
}
void *
crawler_init()
{
GlobalInfo g;
struct itimerspec its;
struct epoll_event ev;
struct epoll_event events[10000];
memset(&g, 0, sizeof(GlobalInfo));
g.transfers = 0;
g.parsed_sites = 0;
g.epfd = epoll_create1(EPOLL_CLOEXEC);
if (g.epfd == -1) {
perror("epoll_create1 failed\n");
exit(1);
}
g.tfd = timerfd_create(CLOCK_MONOTONIC, TFD_NONBLOCK | TFD_CLOEXEC);
if (g.tfd == -1) {
perror("timerfd_create failed\n");
exit(1);
}
memset(&its, 0, sizeof(struct itimerspec));
its.it_interval.tv_sec = 1;
its.it_value.tv_sec = 1;
timerfd_settime(g.tfd, 0, &its, NULL);
ev.events = EPOLLIN;
ev.data.fd = g.tfd;
epoll_ctl(g.epfd, EPOLL_CTL_ADD, g.tfd, &ev);
curl_global_init(CURL_GLOBAL_DEFAULT);
g.multi = curl_multi_init();
/* setup the generic multi interface options we want */
curl_multi_setopt(g.multi, CURLMOPT_SOCKETFUNCTION, sock_cb);
curl_multi_setopt(g.multi, CURLMOPT_SOCKETDATA, &g);
curl_multi_setopt(g.multi, CURLMOPT_TIMERFUNCTION, multi_timer_cb);
curl_multi_setopt(g.multi, CURLMOPT_TIMERDATA, &g);
/* we don't call any curl_multi_socket*() function yet as we have no handles added! */
//printf("Starting crawler...\n");
while (!should_exit) {
int idx;
int err = epoll_wait(g.epfd, events, sizeof(events)/sizeof(struct epoll_event), 10000);
char *url;
url = mysql_url_frontier_pop();
new_head_conn(url, &g);
if (err == -1) {
if (errno == EINTR) {
fprintf(MSG_OUT, "note: wait interrupted\n");
continue;
} else {
perror("epoll_wait");
exit(1);
}
}
for (idx = 0; idx < err; ++idx) {
if (events[idx].data.fd == g.tfd) {
timer_cb(&g, events[idx].events);
} else {
event_cb(&g, events[idx].data.fd, events[idx].events);
}
}
}
fprintf(MSG_OUT, "Exiting normally.\n");
fflush(MSG_OUT);
curl_multi_setopt(g.multi, CURLMOPT_SOCKETFUNCTION, end_sock_cb);
while (g.concurrent_connections > 0 || g.transfers > 0)
{
int idx;
int err = epoll_wait(g.epfd, events, sizeof(events)/sizeof(struct epoll_event), 10000);
if (err == -1) {
if (errno == EINTR) {
fprintf(MSG_OUT, "note: wait interrupted\n");
continue;
} else {
perror("epoll_wait");
exit(1);
}
}
for (idx = 0; idx < err; ++idx) {
if (events[idx].data.fd == g.tfd) {
timer_cb(&g, events[idx].events);
} else {
event_cb(&g, events[idx].data.fd, events[idx].events);
}
}
}
fprintf(MSG_OUT, "Finished all in progress downloads.\n");
fflush(MSG_OUT);
curl_multi_cleanup(g.multi);
curl_global_cleanup();
return (NULL);
}
int
main(int argc, char **argv)
{
int cleanup = 0, opt, ret;
should_exit = 0;
signal(SIGINT, signal_handler);
signal(SIGKILL, signal_handler);
mysql_start();
crawler_init();
mysql_stop();
printf("Exiting.\n");
return (0);
}
典型输出为：
Parsed sites: 42, 6 parallel connections, 4 still running, 6 transfersssExiting normally.
Parsed sites: 48, 0 parallel connections, 0 still running, 0 transfers Finished all in progress downloads.
Exiting.
如您所见，该应用并未达到硬编码的最大并行连接数。差远了。而且它只消耗大约 3 Mbps 的带宽。
为什么？我不明白为什么仅添加 MySQL 代码会导致程序无法 运行 发挥其全部潜力。无法达到硬编码的最大并行连接数。
知道这段代码有什么问题吗？
编辑
这是一些 gprof 输出：
Flat profile:
Each sample counts as 0.01 seconds.
% cumulative self self total
time seconds seconds calls ms/call ms/call name
44.49 0.04 0.04 12668 0.00 0.00 mysql_url_visited_find
33.37 0.07 0.03 12668 0.00 0.00 mysql_url_frontier_push
11.12 0.08 0.01 12668 0.00 0.00 mysql_url_frontier_find
11.12 0.09 0.01 100 0.10 0.90 html_link_find
0.00 0.09 0.00 17355 0.00 0.00 starts_with
0.00 0.09 0.00 12669 0.00 0.00 url_sanitize
0.00 0.09 0.00 2651 0.00 0.00 mcode_or_die
0.00 0.09 0.00 2432 0.00 0.04 check_multi_info
0.00 0.09 0.00 2420 0.00 0.04 event_cb
0.00 0.09 0.00 1288 0.00 0.00 print_progress
0.00 0.09 0.00 425 0.00 0.00 concurrent_connections_dec
0.00 0.09 0.00 425 0.00 0.00 concurrent_connections_inc
0.00 0.09 0.00 425 0.00 0.00 setsock
0.00 0.09 0.00 303 0.00 0.00 remsock
0.00 0.09 0.00 299 0.00 0.00 addsock
0.00 0.09 0.00 219 0.00 0.00 transfers_dec
0.00 0.09 0.00 219 0.00 0.00 transfers_inc
0.00 0.09 0.00 116 0.00 0.00 mysql_url_frontier_pop
0.00 0.09 0.00 116 0.00 0.00 new_head_conn
0.00 0.09 0.00 103 0.00 0.00 new_body_conn
0.00 0.09 0.00 100 0.00 0.90 html_parse
0.00 0.09 0.00 100 0.00 0.00 html_title_find
0.00 0.09 0.00 100 0.00 0.00 mysql_url_visited_push
0.00 0.09 0.00 100 0.00 0.00 parsed_sites_inc
0.00 0.09 0.00 32 0.00 0.01 timer_cb
0.00 0.09 0.00 1 0.00 90.09 crawler_init
0.00 0.09 0.00 1 0.00 0.00 mysql_start
0.00 0.09 0.00 1 0.00 0.00 mysql_stop
您的所有 mysql 代码都是同步的 — 在您等待 mysql 响应查询或确认插入时，您的事件循环没有做任何事情，并且您花费了 90您在 mysql 相关功能中花费的时间百分比。您没有接近并发连接限制，因为数据库瓶颈限制了您发出新 HTTP 请求的速率，并且在您等待D B。
解决这个问题并不简单，但有一些一般性建议：
尽可能避免与数据库对话。 'frontier' 内容看起来像是可以保持在进程中的状态。
使用异步库与数据库交互并将其集成到您的事件循环中，这样您就可以在数据库查询未完成的情况下继续其他工作。
确保你的数据库有必要的索引，或者考虑使用像 redis 这样的东西而不是 mysql，其中访问模式更明确，简单的访问非常快，你可以'真的 "accidentally" 写了一个性能很差的查询。