PHPor 的Blog – 第109页

路由表存储结构及路由查找算法

http://blog.csdn.net/dog250/article/details/6596046

http://blog.c114.net/html/02/157502-59605.html

http://blog.csdn.net/dog250/article/details/5302773

http://basiccoder.com/intro-linux-kernel-hash-rt-1.html

关于swap与缺页

问题：

1. 出现swap换入就意味着缺页吗？

2. 如何查看一个进程的换入换出？

3. 如何演示swap的换入换出？

解答：

1. 需要查一下

参考资料：

http://hi.baidu.com/shichaojiang/item/6fd7aa7950dd8235714423c2

这个写的还不错，一定要看一下

2. 查看进程的换入换出

查看100次，间隔为1s

sar -x pid 1 100

查看子进程的换入换出

sar -X pid 1 100

3. 需要思考一下

缘起：

有这么一些服务器软件（如： tokyotyrant、memcached），他们提供长连接的功能，简单说，server端基本不会主动关闭连接的；非长连接的时候也会存在这种问题，只是长连接时会表现的更明显一下。如果客户端抽风（或者一些恶意的目的），打开连接后，并不关闭而直接离开，这样，就会在server端残留大量的连接，最直接的表象就是： netstat -anpt 时很慢。

真实案例：

具体原因未明，来个命令：

sar -n SOCK ：

11:00:01 AM    totsck    tcpsck    udpsck    rawsck   ip-frag
...
12:20:01 PM       291       198         6         0         0
12:30:01 PM       275       182         6         0         0
12:40:01 PM       284       194         6         0         0
12:50:02 PM       285       192         6         0         0
01:00:01 PM       300       207         7         0         0
01:10:01 PM       299       207         6         0         0
01:30:03 PM    434614    374694         6         0         0
01:50:03 PM    871925    668143         6         0         0
02:00:01 PM    830446    639373         6         0         0
02:10:01 PM    442958    442852         7         0         0
02:20:01 PM    442929    442826         6         0         0
02:30:01 PM    442895    442792         6         0         0
02:40:01 PM    442888    442785         6         0         0
02:50:01 PM    442903    442797         6         0         0
03:00:01 PM    442889    442785         6         0         0
03:10:01 PM    442892    442787         6         0         0
03:20:01 PM    440122    440017         6         0         0
03:30:01 PM    439761    439659         7         0         0
03:40:01 PM    161380    161281         6         0         0
03:50:01 PM       216       118         6         0         0
04:00:01 PM       218       118         6         0         0
04:10:01 PM       222       124         6         0         0

11:00:01 AM totsck tcpsck udpsck rawsck ip-frag

...

12:20:01 PM 291 198 6 0 0

12:30:01 PM 275 182 6 0 0

12:40:01 PM 284 194 6 0 0

12:50:02 PM 285 192 6 0 0

01:00:01 PM 300 207 7 0 0

01:10:01 PM 299 207 6 0 0

01:30:03 PM 434614 374694 6 0 0

01:50:03 PM 871925 668143 6 0 0

02:00:01 PM 830446 639373 6 0 0

02:10:01 PM 442958 442852 7 0 0

02:20:01 PM 442929 442826 6 0 0

02:30:01 PM 442895 442792 6 0 0

02:40:01 PM 442888 442785 6 0 0

02:50:01 PM 442903 442797 6 0 0

03:00:01 PM 442889 442785 6 0 0

03:10:01 PM 442892 442787 6 0 0

03:20:01 PM 440122 440017 6 0 0

03:30:01 PM 439761 439659 7 0 0

03:40:01 PM 161380 161281 6 0 0

03:50:01 PM 216 118 6 0 0

04:00:01 PM 218 118 6 0 0

04:10:01 PM 222 124 6 0 0

我们发现在1:30 PM时，tcp连接数突然飙升到 40万，没有做任何操作3:40 PM时，连接数就开始减少了。

# cat /var/log/messages | grep "port 6101"
Nov 26 13:28:43 localhost kernel: [4771814.322915] possible SYN flooding on port 6101. Sending cookies.
Nov 26 13:32:02 localhost kernel: [4772012.800915] possible SYN flooding on port 6101. Sending cookies.

# cat /var/log/messages | grep "port 6101"

Nov 26 13:28:43 localhost kernel: [4771814.322915] possible SYN flooding on port 6101. Sending cookies.

Nov 26 13:32:02 localhost kernel: [4772012.800915] possible SYN flooding on port 6101. Sending cookies.

原因：

# /sbin/sysctl -a| grep keep
net.ipv4.tcp_keepalive_intvl = 75
net.ipv4.tcp_keepalive_probes = 9
net.ipv4.tcp_keepalive_time = 7200

# /sbin/sysctl -a| grep keep

net.ipv4.tcp_keepalive_intvl = 75

net.ipv4.tcp_keepalive_probes = 9

net.ipv4.tcp_keepalive_time = 7200

连接2小时没有操作，探活机制发现对方已经消失，于是，关闭了连接；看了一下tokyotyrant 和 memcached的源码，都做了KEEPALIVE 的设置：

./ttutil.c:      setsockopt(fd, SOL_SOCKET, SO_KEEPALIVE, (char *)&optint, sizeof(optint));

1	./ttutil.c: setsockopt(fd, SOL_SOCKET, SO_KEEPALIVE, (char *)&optint, sizeof(optint));

./memcached.c:    setsockopt(sfd, SOL_SOCKET, SO_KEEPALIVE, (void *)&flags, sizeof(flags));

1	./memcached.c: setsockopt(sfd, SOL_SOCKET, SO_KEEPALIVE, (void *)&flags, sizeof(flags));

只是都没有做平静时间的设置，于是就都参考系统的设置了。

教训：

1. 如果没有keepalive的机制，系统运行时间过长总会出现这样的问题的

学习：

关于keep-alive的一些参数设置，摘自： man tcp ：

tcp_keepalive_intvl (integer; default: 75)
The number of seconds between TCP keep-alive probes.

如果探活时，对方没有反应则，间隔 tcp_keepalive_intvl 秒后重发探活数据包，避免因丢包产生的探活差错

tcp_keepalive_probes (integer; default: 9)
The maximum number of TCP keep-alive probes to send before giving up and killing the connection if no
response is obtained from the other end.

如果发送了 tcp_keepalive_probes 次探活数据包，对方都没有反应，则视为对方死掉，而不是探活1次失败就放弃

tcp_keepalive_time (integer; default: 7200)
The number of seconds a connection needs to be idle before TCP begins sending out keep-alive probes.
Keep-alives are only sent when the SO_KEEPALIVE socket option is enabled. The default value is 7200 sec-
onds (2 hours). An idle connection is terminated after approximately an additional 11 minutes (9 probes
an interval of 75 seconds apart) when keep-alive is enabled.

Note that underlying connection tracking mechanisms and application timeouts may be much shorter.

从连接平静开始，间隔 tcp_keepalive_time 秒开始探活，避免对方异常退出

编程实现：

使用setsockopt(…) 函数来实现：

  TCP_KEEPALIVE : set idle time in milliseconds
  TCP_KEEPIDLE   : set idle time in seconds
  TCP_KEEPINTVL : set keep alive time interval
  TCP_KEEPCNT    : set number of keep alive probes.

TCP_KEEPALIVE : set idle time in milliseconds

TCP_KEEPIDLE : set idle time in seconds

TCP_KEEPINTVL : set keep alive time interval

TCP_KEEPCNT : set number of keep alive probes.

#define SERV_HOST_ADDR "192.168.1.40"
#define SERV_HOST_PORT 8000

void receiver(void)
{
    int socket_fd, ret;
    struct sockaddr_in serv_addr;
    int num_recv_bytes =0;
    int val=1;
    int keepalive_idle_time = 5000; /* number of milli seconds */

    if ((socket_fd = socket(AF_INET,SOCK_STREAM,0)) == -1)
    {
        return; // error
    }

    memset((char *)&serv_addr,0,sizeof(serv_addr));
    serv_addr.sin_family = AF_INET;
        serv_addr.sin_addr.s_addr = inet_addr(SERV_HOST_ADDR);
        serv_addr.sin_port = htons(SERV_TCP_PORT);

    /* set the idle time after which tcp keepalive probes should start */
        if( setsockopt(socket_fd,IPPROTO_TCP,TCP_KEEPALIVE,(void*)&keepalive_idle_time,sizeof(keepalive_idle_time)) == -1)
    {
        printf("setsockopt TCP_KEEPALIVE failed\n");
        return;
    }

    if( setsockopt(socket_fd,SOL_SOCKET,SO_KEEPALIVE,(void*)&val,sizeof(val)) == -1)
    {
      printf("setsockopt SO_KEEPALIVE failed\n");
      return;
    }

    if (connect(socket_fd,(struct sockaddr *) &serv_addr, sizeof(serv_addr)) <0)
    {
                ret = send(socket_fd,buffer,strlen(buffer),0);
        // Receive the same string, if idle then break
        //
        memset(rcv_buffer,0,sizeof(rcv_buffer));

                while(1)
                {
                        num_recv_bytes = recv(socket_fd,rcv_buffer,BUFFER_LENGTH,0);
                        if(num_recv_bytes < 0)
                        {
                                printf("Receiver: failed exiting\n",buffer);
                                return;
                        }
                        else if (num_recv_bytes == 0)
                        {
                                /*
                                Connection is closed by the tcp keep alive timer upon timeout
                                we will break out of the while loop and close the
                                socket.
                                */
                                printf("Connection IDLE Keep-Alive timer terminated connection\n");
                                break;
                        }
                }

                // close the socket
                close(socket_fd);
        }
}

#define SERV_HOST_ADDR "192.168.1.40"

#define SERV_HOST_PORT 8000

void receiver(void)

{

int socket_fd, ret;

struct sockaddr_in serv_addr;

int num_recv_bytes =0;

int val=1;

int keepalive_idle_time = 5000; /* number of milli seconds */

if ((socket_fd = socket(AF_INET,SOCK_STREAM,0)) == -1)

{

return; // error

}

memset((char *)&serv_addr,0,sizeof(serv_addr));

serv_addr.sin_family = AF_INET;

serv_addr.sin_addr.s_addr = inet_addr(SERV_HOST_ADDR);

serv_addr.sin_port = htons(SERV_TCP_PORT);

/* set the idle time after which tcp keepalive probes should start */

if( setsockopt(socket_fd,IPPROTO_TCP,TCP_KEEPALIVE,(void*)&keepalive_idle_time,sizeof(keepalive_idle_time)) == -1)

{

printf("setsockopt TCP_KEEPALIVE failed\n");

return;

}

if( setsockopt(socket_fd,SOL_SOCKET,SO_KEEPALIVE,(void*)&val,sizeof(val)) == -1)

{

printf("setsockopt SO_KEEPALIVE failed\n");

return;

}

if (connect(socket_fd,(struct sockaddr *) &serv_addr, sizeof(serv_addr)) <0)

{

ret = send(socket_fd,buffer,strlen(buffer),0);

// Receive the same string, if idle then break

memset(rcv_buffer,0,sizeof(rcv_buffer));

while(1)

{

num_recv_bytes = recv(socket_fd,rcv_buffer,BUFFER_LENGTH,0);

if(num_recv_bytes < 0)

{

printf("Receiver: failed exiting\n",buffer);

return;

}

else if (num_recv_bytes == 0)

{

Connection is closed by the tcp keep alive timer upon timeout

we will break out of the while loop and close the

socket.

printf("Connection IDLE Keep-Alive timer terminated connection\n");

break;

}

// close the socket

close(socket_fd);

}

参考资料： http://ez.analog.com/docs/DOC-1862

css_display_block_inline_div_span_li_position_float

display 和 position的用法：

学习一点儿css，应该系统地看一遍的，否则，用到的时候就比较浪费时间了；

曾经总在div的布局上花费时间，今天才发现是因为display属性没有看明白。h1，p，div等元素默认是display:block的，所以，两个div 挨着就总换行； span，a等元素默认display:inline 的，所以，两个span挨着就不会自动换行；现在，了解了 display属性后，就可以方便地控制他们是否换行了。

通过：

通过：

参考资料：

http://www.w3schools.com/css/css_display_visibility.asp

http://www.w3schools.com/css/css_positioning.asp

http://www.w3schools.com/css/css_float.asp

http://www.zhangxinxu.com/wordpress/?p=1565

IN http://www.w3schools.com/css/css_align.asp

Crossbrowser Compatibility Issues

When aligning elements like this, it is always a good idea to predefine margin and padding for the <body> element. This is to avoid visual differences in different browsers.

There is a problem with IE8 and earlier, when using the position property. If a container element (in our case <div>) has a specified width, and the !DOCTYPE declaration is missing, IE8 and earlier versions will add a 17px margin on the right side. This seems to be space reserved for a scrollbar. Always set the !DOCTYPE declaration when using the position property:

从痛苦中得到的教训最深刻

一周来忙个不停，基本都是围绕一个问题：

如果请求执行变慢，web server的进程数迅速飙高，然后就拒绝服务，并且报警，又由于个IDC之间有一些暧昧的关系，于是，一个地方的问题就会迅速导致所有的IDC告急，让人毛骨悚然。

请求执行慢主要有几个方面的问题：

1. 后端存储响应慢

2. 网络出现问题

3. 后端存储死掉

请求执行慢的问题曾经也多次遇到，但是都没有导致过拒绝服务，更没有出现过多个IDC连锁反应的，现在出现这些问题，应该和新环境的web server对进程数的限制太低有很大关系。

当然，系统的设计也存在一些问题。

我想，应该清醒和理智地看待这些天的问题：

1. 新环境存在的问题

2. 系统本身存在的问题

这些天，我遇到了这么多问题，我也深刻地意识到了访问外部资源需要多么的慎重。

另外，这些天的事情使我对监控的理解也有所改变，不是说错了什么，只是原来的理解不够全面和深刻，我需要进步。

我不可能什么都去学习，更不可能什么都能学好，要注意把握一个度的问题，以一个正确的心态来做好自己的事情；既能把工作做好，自己也能有所提高。

如何制作自签名证书

最简单快捷的方法，打开Terminal，使用openssl生成私钥和自签名的x509证书。

openssl req -x509 -out public_key.der -outform der -new -newkey rsa:1024 -keyout private_key.pem -days 3650

1	openssl req -x509 -out public_key.der -outform der -new -newkey rsa:1024 -keyout private_key.pem -days 3650

按照命令行的提示输入内容就行了。

几个说明：

public_key.der是输出的自签名的x509证书，即我们要用的。

private_key.pem是输出的私钥，用来解密的，请妥善保管。

rsa:1024这里的1024是密钥长度，1024是比较安全的，如果需要更安全的话，可以用2048，但是加解密代价也会增加。

-days：证书过期时间，一定要加上这个参数，默认的证书过期时间是30天，一般我们不希望证书这么短就过期，所以写上比较合适的天数，例如这里的3650(10年)。

事实上，这一行命令包含了好几个步骤（我研究下面这些步骤的原因是我手头已经由一个private_key.pem私钥了，想直接用这个来生成x509证书，也就是用到了下面的2-3）

1)创建私钥

openssl genrsa -out private_key.pem 1024

1	openssl genrsa -out private_key.pem 1024

2)创建证书请求（按照提示输入信息）

openssl req -new -out cert.csr -key private_key.pem

1	openssl req -new -out cert.csr -key private_key.pem

3)自签署根证书

openssl x509 -req -in cert.csr -out public_key.der -outform der -signkey private_key.pem -days 3650

1	openssl x509 -req -in cert.csr -out public_key.der -outform der -signkey private_key.pem -days 3650

摘自： http://blog.iamzsx.me/show.html?id=155002

参考资料：

http://blog.csdn.net/jinglijun/article/details/7770301

http://blog.csdn.net/jinglijun/article/details/7770411

PHP 使用证书加密、私钥解密：

<?php
echo "Source: $source";
$fp=fopen("/path/to/certificate.crt","r");
$pub_key=fread($fp,8192);
fclose($fp);
openssl_get_publickey($pub_key);
/*
 * NOTE:  Here you use the $pub_key value (converted, I guess)
 */
openssl_public_encrypt($source,$crypttext,$pub_key);
echo "String crypted: $crypttext";

$fp=fopen("/path/to/private.key","r");
$priv_key=fread($fp,8192);
fclose($fp);
// $passphrase is required if your key is encoded (suggested)
$res = openssl_get_privatekey($priv_key,$passphrase);
/*
 * NOTE:  Here you use the returned resource value
 */
openssl_private_decrypt($crypttext,$newsource,$res);
echo "String decrypt : $newsource";

<?php

echo "Source: $source";

$fp=fopen("/path/to/certificate.crt","r");

$pub_key=fread($fp,8192);

fclose($fp);

openssl_get_publickey($pub_key);

* NOTE: Here you use the $pub_key value (converted, I guess)

openssl_public_encrypt($source,$crypttext,$pub_key);

echo "String crypted: $crypttext";

$fp=fopen("/path/to/private.key","r");

$priv_key=fread($fp,8192);

fclose($fp);

// $passphrase is required if your key is encoded (suggested)

$res = openssl_get_privatekey($priv_key,$passphrase);

* NOTE: Here you use the returned resource value

openssl_private_decrypt($crypttext,$newsource,$res);

echo "String decrypt : $newsource";

PHP连接mongodb

PHP的mongo模块使用域名访问mongodb时，其连接池是用域名做key来记录长连接的

...
  'gm9004.mars.grid.sina.com.cn:9004....28953' => 
  array (
    'in use' => 0,
    'in pool' => 1,
    'remaining' => -2,
    'total' => -1,
    'timeout' => 498,
    'waiting' => 0,
  ),
  'gm9002.mars.grid.sina.com.cn:9002....28953' => 
  array (
    'in use' => 0,
    'in pool' => 1,
    'remaining' => -2,
    'total' => -1,
    'timeout' => 498,
    'waiting' => 0,
  ),
...

...

'gm9004.mars.grid.sina.com.cn:9004....28953' =>

array (

'in use' => 0,

'in pool' => 1,

'remaining' => -2,

'total' => -1,

'timeout' => 498,

'waiting' => 0,

'gm9002.mars.grid.sina.com.cn:9002....28953' =>

array (

'in use' => 0,

'in pool' => 1,

'remaining' => -2,

'total' => -1,

'timeout' => 498,

'waiting' => 0,

...

关于mysql的字符集问题

从认识mysql开始就开始受困于mysql的字符集问题，希望能花点时间搞清楚，下面是一个什么样的问题呢：

“汉字”在utf8的字符集中是6个字节的，然而，这里却是8个字节，怎么解释呢？而且，“汉字”在utf8字符集中的16进制编码应该是“e6 b1 89 e5 ad 97”

参考资料： http://mysql.rjweb.org/doc.php/charcoll

里面有如下一段描述，似乎可以解释上述的问题：

—————————–

“Double encoding” is a term I made up for the following situation. It is a case where “two wrongs make a right”.

⚈  Table defined utf8
⚈  INSERTer declared latin1 (used default instead of doing SET NAMES)
⚈  Data being INSERTed is actually utf8 already.
What happened:
⚈  A 2-byte letter (say, a grave-e) was correctly represented in utf8.
⚈  The INSERT statement handed the 2 bytes to MySQL, but implied the need for conversion
⚈  Each byte was converted to utf8
⚈  The table has 4 bytes.
When SELECTing, the reverse happens 4->2->1, and the user is oblivious of the bug. Looking CHAR_LENGTH will spot it. Strange orderings may happen.

—————————–

看来预期6个字节存储为8个字节还不算严重的，严重的是上面存储翻倍的情况，上面的引用翻译如下：

================

“Double encoding” 是我用来描述下述情况的一个术语，这是一种“错错得对”的情况。

⚈ 表定义为utf8
⚈ INSERT时声明的字符集为latin1（如果没有使用 set names，则默认就是latin1的）
⚈ 要插入的数据确实是utf8编码

发生了什么？
⚈ 一个utf8编码的2字节的字母
⚈ 插入语句写mysql时要对2字节做处理，但是因为字符集声明为latin1的，所以要做编码转换
⚈ 每个字节用2字节的utf8编码表示
⚈ 存储结果为4个字节

在SELECT 的时候，发生一个反向的转换 4->2->1，并且用户不知道这个’bug’,于是就出现了上面的问题

=================

PHP的curl模块的post方法

PHP中curl模块的设置为post后，如果设置post内容为一个数组，则默认并不是使用 Content-Type: application/x-www-form-urlencoded 编码的，如果post内容为 a=A&b=B… 的形式，则使用 Content-Type: application/x-www-form-urlencoded 编码，示例如下：

<?php
$curl = curl_init();

curl_setopt($ch, CURLOPT_POST, 1);
curl_setopt($ch, CURLOPT_POSTFIELDS, $arrQuery);

<?php

$curl = curl_init();

curl_setopt($ch, CURLOPT_POST, 1);

curl_setopt($ch, CURLOPT_POSTFIELDS, $arrQuery);

这时候就不会添加Content-Type: application/x-www-form-urlencoded 头

&lt;?php
$curl = curl_init();

curl_setopt($ch, CURLOPT_POST, 1);
curl_setopt($ch, CURLOPT_POSTFIELDS, http_build_query($arrQuery));

<?php

$curl = curl_init();

curl_setopt($ch, CURLOPT_POST, 1);

curl_setopt($ch, CURLOPT_POSTFIELDS, http_build_query($arrQuery));

这时候就会添加 Content-Type: application/x-www-form-urlencoded 头

也不知道这样的实现是友好还是2

基于事件的网络编程

越来越多的网络软件在采用基于事件的网络编程了，需要详细了解一下事件网络编程的原理和机制。结合memcache、redis的源码了解一下

参考资料：

使用libevent的小例子： http://blog.csdn.net/robertkun/article/details/7885379

libevent示例2：http://blog.sina.com.cn/s/blog_0001988f0100l9ym.html

libevent的使用： http://www.cnblogs.com/cnspace/archive/2011/07/19/2110891.html