Apache_Http_Server MPM: GracefulShutdownTimeout Example
| Description: | A collection of directives that are implemented by more than one multi-processing module (MPM) |
|---|---|
| Status: | MPM |
CoreDumpDirectory Directive
| Description: | Directory where Apache HTTP Server attempts to switch before dumping core |
|---|---|
| Syntax: | CoreDumpDirectory directory |
| Default: | See usage for the default setting |
| Context: | server config |
| Status: | MPM |
| Module: | event, worker, prefork |
This controls the directory to which Apache httpd attempts to switch before dumping core. If your operating system is configured to create core files in the working directory of the crashing process, CoreDumpDirectory is necessary to change working directory from the default ServerRoot directory, which should not be writable by the user the server runs as.
If you want a core dump for debugging, you can use this directive to place it in a different location. This directive has no effect if your operating system is not configured to write core files to the working directory of the crashing processes.
Core Dumps on Linux
If Apache httpd starts as root and switches to another user, the Linux kernel disables core dumps even if the directory is writable for the process. Apache httpd (2.0.46 and later) reenables core dumps on Linux 2.4 and beyond, but only if you explicitly configure a CoreDumpDirectory.
Core Dumps on BSD
To enable core-dumping of suid-executables on BSD-systems (such as FreeBSD), set kern.sugid_coredump to 1.
Specific signals
CoreDumpDirectory processing only occurs for a select set of fatal signals: SIGFPE, SIGILL, SIGABORT, SIGSEGV, and SIGBUS.
On some operating systems, SIGQUIT also results in a core dump but does not go through CoreDumpDirectory or EnableExceptionHook processing, so the core location is dictated entirely by the operating system.
EnableExceptionHook Directive
| Description: | Enables a hook that runs exception handlers after a crash |
|---|---|
| Syntax: | EnableExceptionHook On|Off |
| Default: | EnableExceptionHook Off |
| Context: | server config |
| Status: | MPM |
| Module: | event, worker, prefork |
For safety reasons this directive is only available if the server was configured with the --enable-exception-hook option. It enables a hook that allows external modules to plug in and do something after a child crashed.
There are already two modules, mod_whatkilledus and mod_backtrace that make use of this hook. Please have a look at Jeff Trawick's EnableExceptionHook site for more information about these.
GracefulShutdownTimeout Directive
| Description: | Specify a timeout after which a gracefully shutdown server will exit. |
|---|---|
| Syntax: | GracefulShutdownTimeout seconds |
| Default: | GracefulShutdownTimeout 0 |
| Context: | server config |
| Status: | MPM |
| Module: | event, worker, prefork |
| Compatibility: | Available in version 2.2 and later |
The GracefulShutdownTimeout specifies how many seconds after receiving a "graceful-stop" signal, a server should continue to run, handling the existing connections.
Setting this value to zero means that the server will wait indefinitely until all remaining requests have been fully served.
Listen Directive
| Description: | IP addresses and ports that the server listens to |
|---|---|
| Syntax: | Listen [IP-address:]portnumber [protocol] |
| Context: | server config |
| Status: | MPM |
| Module: | event, worker, prefork, mpm_winnt, mpm_netware, mpmt_os2 |
| Compatibility: | The protocol argument was added in 2.1.5 |
The Listen directive instructs Apache httpd to listen to only specific IP addresses or ports; by default it responds to requests on all IP interfaces. Listen is now a required directive. If it is not in the config file, the server will fail to start. This is a change from previous versions of Apache httpd.
The Listen directive tells the server to accept incoming requests on the specified port or address-and-port combination. If only a port number is specified, the server listens to the given port on all interfaces. If an IP address is given as well as a port, the server will listen on the given port and interface.
Multiple Listen directives may be used to specify a number of addresses and ports to listen to. The server will respond to requests from any of the listed addresses and ports.
For example, to make the server accept connections on both port 80 and port 8000, use:
Listen 80 Listen 8000
To make the server accept connections on two specified interfaces and port numbers, use
Listen 192.170.2.1:80 Listen 192.170.2.5:8000
IPv6 addresses must be surrounded in square brackets, as in the following example:
Listen [2001:db8::a00:20ff:fea7:ccea]:80
The optional protocol argument is not required for most configurations. If not specified, https is the default for port 443 and http the default for all other ports. The protocol is used to determine which module should handle a request, and to apply protocol specific optimizations with the AcceptFilter directive.
You only need to set the protocol if you are running on non-standard ports. For example, running an https site on port 8443:
Listen 192.170.2.1:8443 https
Error condition
MultipleListen directives for the same ip address and port will result in an Address already in use error message. See also
ListenBackLog Directive
| Description: | Maximum length of the queue of pending connections |
|---|---|
| Syntax: | ListenBacklog backlog |
| Default: | ListenBacklog 511 |
| Context: | server config |
| Status: | MPM |
| Module: | event, worker, prefork, mpm_winnt, mpm_netware, mpmt_os2 |
The maximum length of the queue of pending connections. Generally no tuning is needed or desired, however on some systems it is desirable to increase this when under a TCP SYN flood attack. See the backlog parameter to the listen(2) system call.
This will often be limited to a smaller number by the operating system. This varies from OS to OS. Also note that many OSes do not use exactly what is specified as the backlog, but use a number based on (but normally larger than) what is set.
ListenCoresBucketsRatio Directive
| Description: | Ratio between the number of CPU cores (online) and the number of listeners' buckets |
|---|---|
| Syntax: | ListenCoresBucketsRatio ratio |
| Default: | ListenCoresBucketsRatio 0 (disabled) |
| Context: | server config |
| Status: | MPM |
| Module: | event, worker, prefork |
| Compatibility: | Available in Apache HTTP Server 2.4.17, with a kernel supporting the socket option SO_REUSEPORT and distributing new connections evenly across listening processes' (or threads') sockets using it (eg. Linux 3.9 and later, but not the current implementations of SO_REUSEPORT in *BSDs. |
A ratio between the number of (online) CPU cores and the number of listeners' buckets can be used to make Apache HTTP Server create num_cpu_cores / ratio listening buckets, each containing its own Listen-ing socket(s) on the same port(s), and then make each child handle a single bucket (with round-robin distribution of the buckets at children creation time).
Meaning of "online" CPU core
On Linux (and also BSD) a CPU core can be turned on/off if Hotplug is configured, therefore ListenCoresBucketsRatio needs to take this parameter into account while calculating the number of buckets to create.
ListenCoresBucketsRatio can improve the scalability when accepting new connections is/becomes the bottleneck. On systems with a large number of CPU cores, enabling this feature has been tested to show significant performances improvement and shorter responses time.
There must be at least twice the number of CPU cores than the configured ratio for this to be active. The recommended ratio is 8, hence at least 16 cores should be available at runtime when this value is used. The right ratio to obtain maximum performance needs to be calculated for each target system, testing multiple values and observing the variations in your key performance metrics.
This directive influences the calculation of the MinSpareThreads and MaxSpareThreads lower bound values. The number of children processes needs to be a multiple of the number of buckets to optimally accept connections.
Multiple Listeners or Apache HTTP servers on the same IP address and port
Setting the SO_REUSEPORT option on the listening socket(s) consequently allows multiple processes (sharing the same EUID, e.g. root) to bind to the the same IP address and port, without the binding error raised by the system in the usual case.
This also means that multiple instances of Apache httpd configured on a same IP:port and with a positive ListenCoresBucketsRatio would start without an error too, and then run with incoming connections evenly distributed accross both instances (this is NOT a recommendation or a sensible usage in any case, but just a notice that it would prevent such possible issues to be detected).
Within the same instance, Apache httpd will check and fail to start if multiple Listen directives on the exact same IP (or hostname) and port are configured, thus avoiding the creation of some duplicated buckets which would be useless and kill performances. However it can't (and won't try harder to) catch all the possible overlapping cases (like a hostname resolving to an IP used elsewhere).
MaxConnectionsPerChild Directive
| Description: | Limit on the number of connections that an individual child server will handle during its life |
|---|---|
| Syntax: | MaxConnectionsPerChild number |
| Default: | MaxConnectionsPerChild 0 |
| Context: | server config |
| Status: | MPM |
| Module: | event, worker, prefork, mpm_winnt, mpm_netware, mpmt_os2 |
| Compatibility: | Available Apache HTTP Server 2.3.9 and later. The old name MaxRequestsPerChild is still supported. |
The MaxConnectionsPerChild directive sets the limit on the number of connections that an individual child server process will handle. After MaxConnectionsPerChild connections, the child process will die. If MaxConnectionsPerChild is 0, then the process will never expire.
Setting MaxConnectionsPerChild to a non-zero value limits the amount of memory that process can consume by (accidental) memory leakage.
MaxMemFree Directive
| Description: | Maximum amount of memory that the main allocator is allowed to hold without calling free() |
|---|---|
| Syntax: | MaxMemFree KBytes |
| Default: | MaxMemFree 2048 |
| Context: | server config |
| Status: | MPM |
| Module: | event, worker, prefork, mpm_winnt, mpm_netware |
The MaxMemFree directive sets the maximum number of free Kbytes that every allocator is allowed to hold without calling free(). In threaded MPMs, every thread has its own allocator. When set to zero, the threshold will be set to unlimited.
MaxRequestWorkers Directive
| Description: | Maximum number of connections that will be processed simultaneously |
|---|---|
| Syntax: | MaxRequestWorkers number |
| Default: | See usage for details |
| Context: | server config |
| Status: | MPM |
| Module: | event, worker, prefork |
The MaxRequestWorkers directive sets the limit on the number of simultaneous requests that will be served. Any connection attempts over the MaxRequestWorkers limit will normally be queued, up to a number based on the ListenBacklog directive. Once a child process is freed at the end of a different request, the connection will then be serviced.
For non-threaded servers (i.e., prefork), MaxRequestWorkers translates into the maximum number of child processes that will be launched to serve requests. The default value is 256; to increase it, you must also raise ServerLimit.
For threaded and hybrid servers (e.g. event or worker) MaxRequestWorkers restricts the total number of threads that will be available to serve clients. For hybrid MPMs the default value is 16 (ServerLimit) multiplied by the value of 25 (ThreadsPerChild). Therefore, to increase MaxRequestWorkers to a value that requires more than 16 processes, you must also raise ServerLimit.
MaxRequestWorkers was called MaxClients before version 2.3.13. The old name is still supported.
MaxSpareThreads Directive
| Description: | Maximum number of idle threads |
|---|---|
| Syntax: | MaxSpareThreads number |
| Default: | See usage for details |
| Context: | server config |
| Status: | MPM |
| Module: | event, worker, mpm_netware, mpmt_os2 |
Maximum number of idle threads. Different MPMs deal with this directive differently.
For worker and event, the default is MaxSpareThreads 250. These MPMs deal with idle threads on a server-wide basis. If there are too many idle threads in the server then child processes are killed until the number of idle threads is less than this number. Additional processes/threads might be created if ListenCoresBucketsRatio is enabled.
For mpm_netware the default is MaxSpareThreads 100. Since this MPM runs a single-process, the spare thread count is also server-wide.
mpmt_os2 works similar to mpm_netware. For mpmt_os2 the default value is 10.
Restrictions
The range of the MaxSpareThreads value is restricted. Apache httpd will correct the given value automatically according to the following rules:
-
mpm_netwarewants the value to be greater thanMinSpareThreads. - For
workerandevent, the value must be greater or equal to the sum ofMinSpareThreadsandThreadsPerChild.
See also
MinSpareThreads Directive
| Description: | Minimum number of idle threads available to handle request spikes |
|---|---|
| Syntax: | MinSpareThreads number |
| Default: | See usage for details |
| Context: | server config |
| Status: | MPM |
| Module: | event, worker, mpm_netware, mpmt_os2 |
Minimum number of idle threads to handle request spikes. Different MPMs deal with this directive differently.
worker and event use a default of MinSpareThreads 75 and deal with idle threads on a server-wide basis. If there aren't enough idle threads in the server then child processes are created until the number of idle threads is greater than number. Additional processes/threads might be created if ListenCoresBucketsRatio is enabled.
mpm_netware uses a default of MinSpareThreads 10 and, since it is a single-process MPM, tracks this on a server-wide bases.
mpmt_os2 works similar to mpm_netware. For mpmt_os2 the default value is 5.
See also
PidFile Directive
| Description: | File where the server records the process ID of the daemon |
|---|---|
| Syntax: | PidFile filename |
| Default: | PidFile logs/httpd.pid |
| Context: | server config |
| Status: | MPM |
| Module: | event, worker, prefork, mpm_winnt, mpmt_os2 |
The PidFile directive sets the file to which the server records the process id of the daemon. If the filename is not absolute then it is assumed to be relative to the ServerRoot.
Example
PidFile /var/run/apache.pid
It is often useful to be able to send the server a signal, so that it closes and then re-opens its ErrorLog and TransferLog, and re-reads its configuration files. This is done by sending a SIGHUP (kill -1) signal to the process id listed in the PidFile.
The PidFile is subject to the same warnings about log file placement and security.
Note
As of Apache HTTP Server 2, we recommended that you only use the apachectl script, or the init script that your OS provides, for (re-)starting or stopping the server.
ReceiveBufferSize Directive
| Description: | TCP receive buffer size |
|---|---|
| Syntax: | ReceiveBufferSize bytes |
| Default: | ReceiveBufferSize 0 |
| Context: | server config |
| Status: | MPM |
| Module: | event, worker, prefork, mpm_winnt, mpm_netware, mpmt_os2 |
The server will set the TCP receive buffer size to the number of bytes specified.
If set to the value of 0, the server will use the OS default.
ScoreBoardFile Directive
| Description: | Location of the file used to store coordination data for the child processes |
|---|---|
| Syntax: | ScoreBoardFile file-path |
| Default: | ScoreBoardFile logs/apache_runtime_status |
| Context: | server config |
| Status: | MPM |
| Module: | event, worker, prefork, mpm_winnt |
Apache HTTP Server uses a scoreboard to communicate between its parent and child processes. Some architectures require a file to facilitate this communication. If the file is left unspecified, Apache httpd first attempts to create the scoreboard entirely in memory (using anonymous shared memory) and, failing that, will attempt to create the file on disk (using file-based shared memory). Specifying this directive causes Apache httpd to always create the file on the disk.
Example
ScoreBoardFile /var/run/apache_runtime_status
File-based shared memory is useful for third-party applications that require direct access to the scoreboard.
If you use a ScoreBoardFile then you may see improved speed by placing it on a RAM disk. But be careful that you heed the same warnings about log file placement and security.
See also
SendBufferSize Directive
| Description: | TCP buffer size |
|---|---|
| Syntax: | SendBufferSize bytes |
| Default: | SendBufferSize 0 |
| Context: | server config |
| Status: | MPM |
| Module: | event, worker, prefork, mpm_winnt, mpm_netware, mpmt_os2 |
Sets the server's TCP send buffer size to the number of bytes specified. It is often useful to set this past the OS's standard default value on high speed, high latency connections (i.e., 100ms or so, such as transcontinental fast pipes).
If set to the value of 0, the server will use the default value provided by your OS.
Further configuration of your operating system may be required to elicit better performance on high speed, high latency connections.
On some operating systems, changes in TCP behavior resulting from a larger SendBufferSize may not be seen unless EnableSendfile is set to OFF. This interaction applies only to static files.
ServerLimit Directive
| Description: | Upper limit on configurable number of processes |
|---|---|
| Syntax: | ServerLimit number |
| Default: | See usage for details |
| Context: | server config |
| Status: | MPM |
| Module: | event, worker, prefork |
For the prefork MPM, this directive sets the maximum configured value for MaxRequestWorkers for the lifetime of the Apache httpd process. For the worker and event MPMs, this directive in combination with ThreadLimit sets the maximum configured value for MaxRequestWorkers for the lifetime of the Apache httpd process. For the event MPM, this directive also defines how many old server processes may keep running and finish processing open connections. Any attempts to change this directive during a restart will be ignored, but MaxRequestWorkers can be modified during a restart.
Special care must be taken when using this directive. If ServerLimit is set to a value much higher than necessary, extra, unused shared memory will be allocated. If both ServerLimit and MaxRequestWorkers are set to values higher than the system can handle, Apache httpd may not start or the system may become unstable.
With the prefork MPM, use this directive only if you need to set MaxRequestWorkers higher than 256 (default). Do not set the value of this directive any higher than what you might want to set MaxRequestWorkers to.
With worker, use this directive only if your MaxRequestWorkers and ThreadsPerChild settings require more than 16 server processes (default). Do not set the value of this directive any higher than the number of server processes required by what you may want for MaxRequestWorkers and ThreadsPerChild.
With event, increase this directive if the process number defined by your MaxRequestWorkers and ThreadsPerChild settings, plus the number of gracefully shutting down processes, is more than 16 server processes (default).
Note
There is a hard limit of ServerLimit 20000 compiled into the server (for the prefork MPM 200000). This is intended to avoid nasty effects caused by typos. To increase it even further past this limit, you will need to modify the value of MAX_SERVER_LIMIT in the mpm source file and rebuild the server.
See also
StartServers Directive
| Description: | Number of child server processes created at startup |
|---|---|
| Syntax: | StartServers number |
| Default: | See usage for details |
| Context: | server config |
| Status: | MPM |
| Module: | event, worker, prefork, mpmt_os2 |
The StartServers directive sets the number of child server processes created on startup. As the number of processes is dynamically controlled depending on the load, (see MinSpareThreads, MaxSpareThreads, MinSpareServers, MaxSpareServers) there is usually little reason to adjust this parameter.
The default value differs from MPM to MPM. worker and event default to StartServers 3; prefork defaults to 5; mpmt_os2 defaults to 2.
StartThreads Directive
| Description: | Number of threads created on startup |
|---|---|
| Syntax: | StartThreads number |
| Default: | See usage for details |
| Context: | server config |
| Status: | MPM |
| Module: | mpm_netware |
Number of threads created on startup. As the number of threads is dynamically controlled depending on the load, (see MinSpareThreads, MaxSpareThreads, MinSpareServers, MaxSpareServers) there is usually little reason to adjust this parameter.
For mpm_netware the default is StartThreads 50 and, since there is only a single process, this is the total number of threads created at startup to serve requests.
ThreadLimit Directive
| Description: | Sets the upper limit on the configurable number of threads per child process |
|---|---|
| Syntax: | ThreadLimit number |
| Default: | See usage for details |
| Context: | server config |
| Status: | MPM |
| Module: | event, worker, mpm_winnt |
This directive sets the maximum configured value for ThreadsPerChild for the lifetime of the Apache httpd process. Any attempts to change this directive during a restart will be ignored, but ThreadsPerChild can be modified during a restart up to the value of this directive.
Special care must be taken when using this directive. If ThreadLimit is set to a value much higher than ThreadsPerChild, extra unused shared memory will be allocated. If both ThreadLimit and ThreadsPerChild are set to values higher than the system can handle, Apache httpd may not start or the system may become unstable. Do not set the value of this directive any higher than your greatest predicted setting of ThreadsPerChild for the current run of Apache httpd.
The default value for ThreadLimit is 1920 when used with mpm_winnt and 64 when used with the others.
Note
There is a hard limit of ThreadLimit 20000 (or ThreadLimit 100000 with event, ThreadLimit 15000 with mpm_winnt) compiled into the server. This is intended to avoid nasty effects caused by typos. To increase it even further past this limit, you will need to modify the value of MAX_THREAD_LIMIT in the mpm source file and rebuild the server.
ThreadsPerChild Directive
| Description: | Number of threads created by each child process |
|---|---|
| Syntax: | ThreadsPerChild number |
| Default: | See usage for details |
| Context: | server config |
| Status: | MPM |
| Module: | event, worker, mpm_winnt |
This directive sets the number of threads created by each child process. The child creates these threads at startup and never creates more. If using an MPM like mpm_winnt, where there is only one child process, this number should be high enough to handle the entire load of the server. If using an MPM like worker, where there are multiple child processes, the total number of threads should be high enough to handle the common load on the server.
The default value for ThreadsPerChild is 64 when used with mpm_winnt and 25 when used with the others.
ThreadStackSize Directive
| Description: | The size in bytes of the stack used by threads handling client connections |
|---|---|
| Syntax: | ThreadStackSize size |
| Default: | 65536 on NetWare; varies on other operating systems |
| Context: | server config |
| Status: | MPM |
| Module: | event, worker, mpm_winnt, mpm_netware, mpmt_os2 |
| Compatibility: | Available in Apache HTTP Server 2.1 and later |
The ThreadStackSize directive sets the size of the stack (for autodata) of threads which handle client connections and call modules to help process those connections. In most cases the operating system default for stack size is reasonable, but there are some conditions where it may need to be adjusted:
- On platforms with a relatively small default thread stack size (e.g., HP-UX), Apache httpd may crash when using some third-party modules which use a relatively large amount of autodata storage. Those same modules may have worked fine on other platforms where the default thread stack size is larger. This type of crash is resolved by setting
ThreadStackSizeto a value higher than the operating system default. This type of adjustment is necessary only if the provider of the third-party module specifies that it is required, or if diagnosis of an Apache httpd crash indicates that the thread stack size was too small. - On platforms where the default thread stack size is significantly larger than necessary for the web server configuration, a higher number of threads per child process will be achievable if
ThreadStackSizeis set to a value lower than the operating system default. This type of adjustment should only be made in a test environment which allows the full set of web server processing can be exercised, as there may be infrequent requests which require more stack to process. The minimum required stack size strongly depends on the modules used, but any change in the web server configuration can invalidate the currentThreadStackSizesetting. - On Linux, this directive can only be used to increase the default stack size, as the underlying system call uses the value as a minimum stack size. The (often large) soft limit for
ulimit -s(8MB if unlimited) is used as the default stack size.
ThreadStackSize unless a high number of threads per child process is needed. On some platforms (including Linux), a setting of 128000 is already too low and causes crashes with some common modules.
Comments
Post a Comment