<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd"> <HTML ><HEAD ><TITLE >Managing Kernel Resources</TITLE ><META NAME="GENERATOR" CONTENT="Modular DocBook HTML Stylesheet Version 1.79"><LINK REV="MADE" HREF="mailto:pgsql-docs@postgresql.org"><LINK REL="HOME" TITLE="PostgreSQL 8.4.12 Documentation" HREF="index.html"><LINK REL="UP" TITLE="Server Setup and Operation" HREF="runtime.html"><LINK REL="PREVIOUS" TITLE="Starting the Database Server" HREF="server-start.html"><LINK REL="NEXT" TITLE="Shutting Down the Server" HREF="server-shutdown.html"><LINK REL="STYLESHEET" TYPE="text/css" HREF="stylesheet.css"><META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=ISO-8859-1"><META NAME="creation" CONTENT="2012-05-31T23:30:11"></HEAD ><BODY CLASS="SECT1" ><DIV CLASS="NAVHEADER" ><TABLE SUMMARY="Header navigation table" WIDTH="100%" BORDER="0" CELLPADDING="0" CELLSPACING="0" ><TR ><TH COLSPAN="5" ALIGN="center" VALIGN="bottom" >PostgreSQL 8.4.12 Documentation</TH ></TR ><TR ><TD WIDTH="10%" ALIGN="left" VALIGN="top" ><A HREF="server-start.html" ACCESSKEY="P" >Prev</A ></TD ><TD WIDTH="10%" ALIGN="left" VALIGN="top" ><A HREF="runtime.html" >Fast Backward</A ></TD ><TD WIDTH="60%" ALIGN="center" VALIGN="bottom" >Chapter 17. Server Setup and Operation</TD ><TD WIDTH="10%" ALIGN="right" VALIGN="top" ><A HREF="runtime.html" >Fast Forward</A ></TD ><TD WIDTH="10%" ALIGN="right" VALIGN="top" ><A HREF="server-shutdown.html" ACCESSKEY="N" >Next</A ></TD ></TR ></TABLE ><HR ALIGN="LEFT" WIDTH="100%"></DIV ><DIV CLASS="SECT1" ><H1 CLASS="SECT1" ><A NAME="KERNEL-RESOURCES" >17.4. Managing Kernel Resources</A ></H1 ><P > A large <SPAN CLASS="PRODUCTNAME" >PostgreSQL</SPAN > installation can quickly exhaust various operating system resource limits. (On some systems, the factory defaults are so low that you don't even need a really <SPAN CLASS="QUOTE" >"large"</SPAN > installation.) If you have encountered this kind of problem, keep reading. </P ><DIV CLASS="SECT2" ><H2 CLASS="SECT2" ><A NAME="SYSVIPC" >17.4.1. Shared Memory and Semaphores</A ></H2 ><A NAME="AEN23755" ></A ><A NAME="AEN23757" ></A ><P > Shared memory and semaphores are collectively referred to as <SPAN CLASS="QUOTE" >"<SPAN CLASS="SYSTEMITEM" >System V</SPAN > <ACRONYM CLASS="ACRONYM" >IPC</ACRONYM >"</SPAN > (together with message queues, which are not relevant for <SPAN CLASS="PRODUCTNAME" >PostgreSQL</SPAN >). Almost all modern operating systems provide these features, but not all of them have them turned on or sufficiently sized by default, especially systems with BSD heritage. (On <SPAN CLASS="SYSTEMITEM" >Windows</SPAN >, <SPAN CLASS="PRODUCTNAME" >PostgreSQL</SPAN > provides its own replacement implementation of these facilities, and so most of this section can be disregarded.) </P ><P > The complete lack of these facilities is usually manifested by an <SPAN CLASS="ERRORNAME" >Illegal system call</SPAN > error upon server start. In that case there's nothing left to do but to reconfigure your kernel. <SPAN CLASS="PRODUCTNAME" >PostgreSQL</SPAN > won't work without them. </P ><P > When <SPAN CLASS="PRODUCTNAME" >PostgreSQL</SPAN > exceeds one of the various hard <ACRONYM CLASS="ACRONYM" >IPC</ACRONYM > limits, the server will refuse to start and should leave an instructive error message describing the problem encountered and what to do about it. (See also <A HREF="server-start.html#SERVER-START-FAILURES" >Section 17.3.1</A >.) The relevant kernel parameters are named consistently across different systems; <A HREF="kernel-resources.html#SYSVIPC-PARAMETERS" >Table 17-1</A > gives an overview. The methods to set them, however, vary. Suggestions for some platforms are given below. Be warned that it is often necessary to reboot your machine, and possibly even recompile the kernel, to change these settings. </P ><DIV CLASS="TABLE" ><A NAME="SYSVIPC-PARAMETERS" ></A ><P ><B >Table 17-1. <SPAN CLASS="SYSTEMITEM" >System V</SPAN > <ACRONYM CLASS="ACRONYM" >IPC</ACRONYM > parameters</B ></P ><TABLE BORDER="1" CLASS="CALSTABLE" ><COL><COL><COL><THEAD ><TR ><TH >Name</TH ><TH >Description</TH ><TH >Reasonable values</TH ></TR ></THEAD ><TBODY ><TR ><TD ><TT CLASS="VARNAME" >SHMMAX</TT ></TD ><TD >Maximum size of shared memory segment (bytes)</TD ><TD >at least several megabytes (see text)</TD ></TR ><TR ><TD ><TT CLASS="VARNAME" >SHMMIN</TT ></TD ><TD >Minimum size of shared memory segment (bytes)</TD ><TD >1</TD ></TR ><TR ><TD ><TT CLASS="VARNAME" >SHMALL</TT ></TD ><TD >Total amount of shared memory available (bytes or pages)</TD ><TD >if bytes, same as <TT CLASS="VARNAME" >SHMMAX</TT >; if pages, <TT CLASS="LITERAL" >ceil(SHMMAX/PAGE_SIZE)</TT ></TD ></TR ><TR ><TD ><TT CLASS="VARNAME" >SHMSEG</TT ></TD ><TD >Maximum number of shared memory segments per process</TD ><TD >only 1 segment is needed, but the default is much higher</TD ></TR ><TR ><TD ><TT CLASS="VARNAME" >SHMMNI</TT ></TD ><TD >Maximum number of shared memory segments system-wide</TD ><TD >like <TT CLASS="VARNAME" >SHMSEG</TT > plus room for other applications</TD ></TR ><TR ><TD ><TT CLASS="VARNAME" >SEMMNI</TT ></TD ><TD >Maximum number of semaphore identifiers (i.e., sets)</TD ><TD >at least <TT CLASS="LITERAL" >ceil((max_connections + autovacuum_max_workers) / 16)</TT ></TD ></TR ><TR ><TD ><TT CLASS="VARNAME" >SEMMNS</TT ></TD ><TD >Maximum number of semaphores system-wide</TD ><TD ><TT CLASS="LITERAL" >ceil((max_connections + autovacuum_max_workers) / 16) * 17</TT > plus room for other applications</TD ></TR ><TR ><TD ><TT CLASS="VARNAME" >SEMMSL</TT ></TD ><TD >Maximum number of semaphores per set</TD ><TD >at least 17</TD ></TR ><TR ><TD ><TT CLASS="VARNAME" >SEMMAP</TT ></TD ><TD >Number of entries in semaphore map</TD ><TD >see text</TD ></TR ><TR ><TD ><TT CLASS="VARNAME" >SEMVMX</TT ></TD ><TD >Maximum value of semaphore</TD ><TD >at least 1000 (The default is often 32767, don't change unless forced to)</TD ></TR ></TBODY ></TABLE ></DIV ><P > <A NAME="AEN23841" ></A > The most important shared memory parameter is <TT CLASS="VARNAME" >SHMMAX</TT >, the maximum size, in bytes, of a shared memory segment. If you get an error message from <CODE CLASS="FUNCTION" >shmget</CODE > like <SPAN CLASS="ERRORNAME" >Invalid argument</SPAN >, it is likely that this limit has been exceeded. The size of the required shared memory segment varies depending on several <SPAN CLASS="PRODUCTNAME" >PostgreSQL</SPAN > configuration parameters, as shown in <A HREF="kernel-resources.html#SHARED-MEMORY-PARAMETERS" >Table 17-2</A >. (Any error message you might get will include the exact size of the failed allocation request.) You can, as a temporary solution, lower some of those settings to avoid the failure. While it is possible to get <SPAN CLASS="PRODUCTNAME" >PostgreSQL</SPAN > to run with <TT CLASS="VARNAME" >SHMMAX</TT > as small as 2 MB, you need considerably more for acceptable performance. Desirable settings are in the tens to hundreds of megabytes. </P ><P > Some systems also have a limit on the total amount of shared memory in the system (<TT CLASS="VARNAME" >SHMALL</TT >). Make sure this is large enough for <SPAN CLASS="PRODUCTNAME" >PostgreSQL</SPAN > plus any other applications that are using shared memory segments. (Caution: <TT CLASS="VARNAME" >SHMALL</TT > is measured in pages rather than bytes on many systems.) </P ><P > Less likely to cause problems is the minimum size for shared memory segments (<TT CLASS="VARNAME" >SHMMIN</TT >), which should be at most approximately 500 kB for <SPAN CLASS="PRODUCTNAME" >PostgreSQL</SPAN > (it is usually just 1). The maximum number of segments system-wide (<TT CLASS="VARNAME" >SHMMNI</TT >) or per-process (<TT CLASS="VARNAME" >SHMSEG</TT >) are unlikely to cause a problem unless your system has them set to zero. </P ><P > <SPAN CLASS="PRODUCTNAME" >PostgreSQL</SPAN > uses one semaphore per allowed connection (<A HREF="runtime-config-connection.html#GUC-MAX-CONNECTIONS" >max_connections</A >) and allowed autovacuum worker process (<A HREF="runtime-config-autovacuum.html#GUC-AUTOVACUUM-MAX-WORKERS" >autovacuum_max_workers</A >), in sets of 16. Each such set will also contain a 17th semaphore which contains a <SPAN CLASS="QUOTE" >"magic number"</SPAN >, to detect collision with semaphore sets used by other applications. The maximum number of semaphores in the system is set by <TT CLASS="VARNAME" >SEMMNS</TT >, which consequently must be at least as high as <TT CLASS="VARNAME" >max_connections</TT > plus <TT CLASS="VARNAME" >autovacuum_max_workers</TT >, plus one extra for each 16 allowed connections plus workers (see the formula in <A HREF="kernel-resources.html#SYSVIPC-PARAMETERS" >Table 17-1</A >). The parameter <TT CLASS="VARNAME" >SEMMNI</TT > determines the limit on the number of semaphore sets that can exist on the system at one time. Hence this parameter must be at least <TT CLASS="LITERAL" >ceil((max_connections + autovacuum_max_workers) / 16)</TT >. Lowering the number of allowed connections is a temporary workaround for failures, which are usually confusingly worded <SPAN CLASS="ERRORNAME" >No space left on device</SPAN >, from the function <CODE CLASS="FUNCTION" >semget</CODE >. </P ><P > In some cases it might also be necessary to increase <TT CLASS="VARNAME" >SEMMAP</TT > to be at least on the order of <TT CLASS="VARNAME" >SEMMNS</TT >. This parameter defines the size of the semaphore resource map, in which each contiguous block of available semaphores needs an entry. When a semaphore set is freed it is either added to an existing entry that is adjacent to the freed block or it is registered under a new map entry. If the map is full, the freed semaphores get lost (until reboot). Fragmentation of the semaphore space could over time lead to fewer available semaphores than there should be. </P ><P > The <TT CLASS="VARNAME" >SEMMSL</TT > parameter, which determines how many semaphores can be in a set, must be at least 17 for <SPAN CLASS="PRODUCTNAME" >PostgreSQL</SPAN >. </P ><P > Various other settings related to <SPAN CLASS="QUOTE" >"semaphore undo"</SPAN >, such as <TT CLASS="VARNAME" >SEMMNU</TT > and <TT CLASS="VARNAME" >SEMUME</TT >, are not of concern for <SPAN CLASS="PRODUCTNAME" >PostgreSQL</SPAN >. </P ><P ></P ><DIV CLASS="VARIABLELIST" ><DL ><DT ><SPAN CLASS="SYSTEMITEM" >AIX</SPAN ></DT ><DD ><P > At least as of version 5.1, it should not be necessary to do any special configuration for such parameters as <TT CLASS="VARNAME" >SHMMAX</TT >, as it appears this is configured to allow all memory to be used as shared memory. That is the sort of configuration commonly used for other databases such as <SPAN CLASS="APPLICATION" >DB/2</SPAN >.</P ><P > It might, however, be necessary to modify the global <TT CLASS="COMMAND" >ulimit</TT > information in <TT CLASS="FILENAME" >/etc/security/limits</TT >, as the default hard limits for file sizes (<TT CLASS="VARNAME" >fsize</TT >) and numbers of files (<TT CLASS="VARNAME" >nofiles</TT >) might be too low. </P ></DD ><DT ><SPAN CLASS="SYSTEMITEM" >BSD/OS</SPAN ></DT ><DD ><DIV CLASS="FORMALPARA" ><P ><B >Shared Memory. </B > By default, only 4 MB of shared memory is supported. Keep in mind that shared memory is not pageable; it is locked in RAM. To increase the amount of shared memory supported by your system, add something like the following to your kernel configuration file: </P><PRE CLASS="PROGRAMLISTING" >options "SHMALL=8192" options "SHMMAX=\(SHMALL*PAGE_SIZE\)"</PRE ><P> <TT CLASS="VARNAME" >SHMALL</TT > is measured in 4 kB pages, so a value of 1024 represents 4 MB of shared memory. Therefore the above increases the maximum shared memory area to 32 MB. For those running 4.3 or later, you will probably also need to increase <TT CLASS="VARNAME" >KERNEL_VIRTUAL_MB</TT > above the default <TT CLASS="LITERAL" >248</TT >. Once all changes have been made, recompile the kernel, and reboot. </P ></DIV ><P > For those running 4.0 and earlier releases, use <TT CLASS="COMMAND" >bpatch</TT > to find the <TT CLASS="VARNAME" >sysptsize</TT > value in the current kernel. This is computed dynamically at boot time. </P><PRE CLASS="SCREEN" >$ <KBD CLASS="USERINPUT" >bpatch -r sysptsize</KBD > <SAMP CLASS="COMPUTEROUTPUT" >0x9 = 9</SAMP ></PRE ><P> Next, add <TT CLASS="VARNAME" >SYSPTSIZE</TT > as a hard-coded value in the kernel configuration file. Increase the value you found using <TT CLASS="COMMAND" >bpatch</TT >. Add 1 for every additional 4 MB of shared memory you desire. </P><PRE CLASS="PROGRAMLISTING" >options "SYSPTSIZE=16"</PRE ><P> <TT CLASS="VARNAME" >sysptsize</TT > cannot be changed by <TT CLASS="COMMAND" >sysctl</TT >. </P ><DIV CLASS="FORMALPARA" ><P ><B >Semaphores. </B > You will probably want to increase the number of semaphores as well; the default system total of 60 will only allow about 50 <SPAN CLASS="PRODUCTNAME" >PostgreSQL</SPAN > connections. Set the values you want in your kernel configuration file, e.g.: </P><PRE CLASS="PROGRAMLISTING" >options "SEMMNI=40" options "SEMMNS=240"</PRE ><P> </P ></DIV ></DD ><DT ><SPAN CLASS="SYSTEMITEM" >FreeBSD</SPAN ></DT ><DD ><P > The default settings are only suitable for small installations (for example, default <TT CLASS="VARNAME" >SHMMAX</TT > is 32 MB). Changes can be made via the <TT CLASS="COMMAND" >sysctl</TT > or <TT CLASS="COMMAND" >loader</TT > interfaces. The following parameters can be set using <TT CLASS="COMMAND" >sysctl</TT >: </P><PRE CLASS="SCREEN" ><SAMP CLASS="PROMPT" >$</SAMP > <KBD CLASS="USERINPUT" >sysctl -w kern.ipc.shmall=32768</KBD > <SAMP CLASS="PROMPT" >$</SAMP > <KBD CLASS="USERINPUT" >sysctl -w kern.ipc.shmmax=134217728</KBD > <SAMP CLASS="PROMPT" >$</SAMP > <KBD CLASS="USERINPUT" >sysctl -w kern.ipc.semmap=256</KBD ></PRE ><P> To have these settings persist over reboots, modify <TT CLASS="FILENAME" >/etc/sysctl.conf</TT >. </P ><P > The remaining semaphore settings are read-only as far as <TT CLASS="COMMAND" >sysctl</TT > is concerned, but can be changed before boot using the <TT CLASS="COMMAND" >loader</TT > prompt: </P><PRE CLASS="SCREEN" ><SAMP CLASS="PROMPT" >(loader)</SAMP > <KBD CLASS="USERINPUT" >set kern.ipc.semmni=256</KBD > <SAMP CLASS="PROMPT" >(loader)</SAMP > <KBD CLASS="USERINPUT" >set kern.ipc.semmns=512</KBD > <SAMP CLASS="PROMPT" >(loader)</SAMP > <KBD CLASS="USERINPUT" >set kern.ipc.semmnu=256</KBD ></PRE ><P> Similarly these can be saved between reboots in <TT CLASS="FILENAME" >/boot/loader.conf</TT >. </P ><P > You might also want to configure your kernel to lock shared memory into RAM and prevent it from being paged out to swap. This can be accomplished using the <TT CLASS="COMMAND" >sysctl</TT > setting <TT CLASS="LITERAL" >kern.ipc.shm_use_phys</TT >. </P ><P > If running in FreeBSD jails by enabling <SPAN CLASS="APPLICATION" >sysctl</SPAN >'s <TT CLASS="LITERAL" >security.jail.sysvipc_allowed</TT >, <SPAN CLASS="APPLICATION" >postmaster</SPAN >s running in different jails should be run by different operating system users. This improves security because it prevents non-root users from interfering with shared memory or semaphores in a different jail, and it allows the PostgreSQL IPC cleanup code to function properly. (In FreeBSD 6.0 and later the IPC cleanup code doesn't properly detect processes in other jails, preventing the running of postmasters on the same port in different jails.) </P ><P > <SPAN CLASS="SYSTEMITEM" >FreeBSD</SPAN > versions before 4.0 work like <SPAN CLASS="SYSTEMITEM" >OpenBSD</SPAN > (see below). </P ></DD ><DT ><SPAN CLASS="SYSTEMITEM" >NetBSD</SPAN ></DT ><DD ><P > In <SPAN CLASS="SYSTEMITEM" >NetBSD</SPAN > 5.0 and later, IPC parameters can be adjusted using <TT CLASS="COMMAND" >sysctl</TT >, for example: </P><PRE CLASS="SCREEN" ><SAMP CLASS="PROMPT" >$</SAMP > <KBD CLASS="USERINPUT" >sysctl -w kern.ipc.shmmax=16777216</KBD ></PRE ><P> To have these settings persist over reboots, modify <TT CLASS="FILENAME" >/etc/sysctl.conf</TT >. </P ><P > You might also want to configure your kernel to lock shared memory into RAM and prevent it from being paged out to swap. This can be accomplished using the <TT CLASS="COMMAND" >sysctl</TT > setting <TT CLASS="LITERAL" >kern.ipc.shm_use_phys</TT >. </P ><P > <SPAN CLASS="SYSTEMITEM" >NetBSD</SPAN > versions before 5.0 work like <SPAN CLASS="SYSTEMITEM" >OpenBSD</SPAN > (see below), except that parameters should be set with the keyword <TT CLASS="LITERAL" >options</TT > not <TT CLASS="LITERAL" >option</TT >. </P ></DD ><DT ><SPAN CLASS="SYSTEMITEM" >OpenBSD</SPAN ></DT ><DD ><P > The options <TT CLASS="VARNAME" >SYSVSHM</TT > and <TT CLASS="VARNAME" >SYSVSEM</TT > need to be enabled when the kernel is compiled. (They are by default.) The maximum size of shared memory is determined by the option <TT CLASS="VARNAME" >SHMMAXPGS</TT > (in pages). The following shows an example of how to set the various parameters: </P><PRE CLASS="PROGRAMLISTING" >option SYSVSHM option SHMMAXPGS=4096 option SHMSEG=256 option SYSVSEM option SEMMNI=256 option SEMMNS=512 option SEMMNU=256 option SEMMAP=256</PRE ><P> </P ><P > You might also want to configure your kernel to lock shared memory into RAM and prevent it from being paged out to swap. This can be accomplished using the <TT CLASS="COMMAND" >sysctl</TT > setting <TT CLASS="LITERAL" >kern.ipc.shm_use_phys</TT >. </P ></DD ><DT ><SPAN CLASS="SYSTEMITEM" >HP-UX</SPAN ></DT ><DD ><P > The default settings tend to suffice for normal installations. On <SPAN CLASS="PRODUCTNAME" >HP-UX</SPAN > 10, the factory default for <TT CLASS="VARNAME" >SEMMNS</TT > is 128, which might be too low for larger database sites. </P ><P > <ACRONYM CLASS="ACRONYM" >IPC</ACRONYM > parameters can be set in the <SPAN CLASS="APPLICATION" >System Administration Manager</SPAN > (<ACRONYM CLASS="ACRONYM" >SAM</ACRONYM >) under <SPAN CLASS="GUIMENU" >Kernel Configuration</SPAN >-><SPAN CLASS="GUIMENUITEM" >Configurable Parameters</SPAN >. Hit <SPAN CLASS="GUIBUTTON" >Create A New Kernel</SPAN > when you're done. </P ></DD ><DT ><SPAN CLASS="SYSTEMITEM" >Linux</SPAN ></DT ><DD ><P > The default maximum segment size is 32 MB, which is only adequate for small <SPAN CLASS="PRODUCTNAME" >PostgreSQL</SPAN > installations. However, the remaining defaults are quite generously sized, and usually do not require changes. The maximum shared memory segment size can be changed via the <TT CLASS="COMMAND" >sysctl</TT > interface. For example, to allow 128 MB, and explicitly set the maximum total shared memory size to 2097152 pages (the default): </P><PRE CLASS="SCREEN" ><SAMP CLASS="PROMPT" >$</SAMP > <KBD CLASS="USERINPUT" >sysctl -w kernel.shmmax=134217728</KBD > <SAMP CLASS="PROMPT" >$</SAMP > <KBD CLASS="USERINPUT" >sysctl -w kernel.shmall=2097152</KBD ></PRE ><P> In addition these settings can be saved between reboots in <TT CLASS="FILENAME" >/etc/sysctl.conf</TT >. </P ><P > Older distributions might not have the <TT CLASS="COMMAND" >sysctl</TT > program, but equivalent changes can be made by manipulating the <TT CLASS="FILENAME" >/proc</TT > file system: </P><PRE CLASS="SCREEN" ><SAMP CLASS="PROMPT" >$</SAMP > <KBD CLASS="USERINPUT" >echo 134217728 >/proc/sys/kernel/shmmax</KBD > <SAMP CLASS="PROMPT" >$</SAMP > <KBD CLASS="USERINPUT" >echo 2097152 >/proc/sys/kernel/shmall</KBD ></PRE ><P> </P ></DD ><DT ><SPAN CLASS="SYSTEMITEM" >MacOS X</SPAN ></DT ><DD ><P > In OS X 10.2 and earlier, edit the file <TT CLASS="FILENAME" >/System/Library/StartupItems/SystemTuning/SystemTuning</TT > and change the values in the following commands: </P><PRE CLASS="PROGRAMLISTING" >sysctl -w kern.sysv.shmmax sysctl -w kern.sysv.shmmin sysctl -w kern.sysv.shmmni sysctl -w kern.sysv.shmseg sysctl -w kern.sysv.shmall</PRE ><P> </P ><P > In OS X 10.3 and later, these commands have been moved to <TT CLASS="FILENAME" >/etc/rc</TT > and must be edited there. Note that <TT CLASS="FILENAME" >/etc/rc</TT > is usually overwritten by OS X updates (such as 10.3.6 to 10.3.7) so you should expect to have to redo your editing after each update. </P ><P > In OS X 10.3.9 and later, instead of editing <TT CLASS="FILENAME" >/etc/rc</TT > you can create a file named <TT CLASS="FILENAME" >/etc/sysctl.conf</TT >, containing variable assignments such as: </P><PRE CLASS="PROGRAMLISTING" >kern.sysv.shmmax=4194304 kern.sysv.shmmin=1 kern.sysv.shmmni=32 kern.sysv.shmseg=8 kern.sysv.shmall=1024</PRE ><P> This method is better than editing <TT CLASS="FILENAME" >/etc/rc</TT > because your changes will be preserved across system updates. Note that <SPAN CLASS="emphasis" ><I CLASS="EMPHASIS" >all five</I ></SPAN > shared-memory parameters must be set in <TT CLASS="FILENAME" >/etc/sysctl.conf</TT >, else the values will be ignored. </P ><P > Beware that recent releases of OS X ignore attempts to set <TT CLASS="VARNAME" >SHMMAX</TT > to a value that isn't an exact multiple of 4096. </P ><P > <TT CLASS="VARNAME" >SHMALL</TT > is measured in 4 kB pages on this platform. </P ><P > In all OS X versions, you'll need to reboot to make changes in the shared memory parameters take effect. </P ></DD ><DT ><SPAN CLASS="SYSTEMITEM" >SCO OpenServer</SPAN ></DT ><DD ><P > In the default configuration, only 512 kB of shared memory per segment is allowed. To increase the setting, first change to the directory <TT CLASS="FILENAME" >/etc/conf/cf.d</TT >. To display the current value of <TT CLASS="VARNAME" >SHMMAX</TT >, run: </P><PRE CLASS="PROGRAMLISTING" >./configure -y SHMMAX</PRE ><P> To set a new value for <TT CLASS="VARNAME" >SHMMAX</TT >, run: </P><PRE CLASS="PROGRAMLISTING" >./configure SHMMAX=<TT CLASS="REPLACEABLE" ><I >value</I ></TT ></PRE ><P> where <TT CLASS="REPLACEABLE" ><I >value</I ></TT > is the new value you want to use (in bytes). After setting <TT CLASS="VARNAME" >SHMMAX</TT >, rebuild the kernel: </P><PRE CLASS="PROGRAMLISTING" >./link_unix</PRE ><P> and reboot. </P ></DD ><DT ><SPAN CLASS="SYSTEMITEM" >Solaris</SPAN ></DT ><DD ><P > At least in version 2.6, the default maximum size of a shared memory segment is too low for <SPAN CLASS="PRODUCTNAME" >PostgreSQL</SPAN >. The relevant settings can be changed in <TT CLASS="FILENAME" >/etc/system</TT >, for example: </P><PRE CLASS="PROGRAMLISTING" >set shmsys:shminfo_shmmax=0x2000000 set shmsys:shminfo_shmmin=1 set shmsys:shminfo_shmmni=256 set shmsys:shminfo_shmseg=256 set semsys:seminfo_semmap=256 set semsys:seminfo_semmni=512 set semsys:seminfo_semmns=512 set semsys:seminfo_semmsl=32</PRE ><P> You need to reboot for the changes to take effect. </P ><P > See also <A HREF="http://sunsite.uakom.sk/sunworldonline/swol-09-1997/swol-09-insidesolaris.html" TARGET="_top" >http://sunsite.uakom.sk/sunworldonline/swol-09-1997/swol-09-insidesolaris.html</A > for information on shared memory under <SPAN CLASS="PRODUCTNAME" >Solaris</SPAN >. </P ></DD ><DT ><SPAN CLASS="SYSTEMITEM" >UnixWare</SPAN ></DT ><DD ><P > On <SPAN CLASS="PRODUCTNAME" >UnixWare</SPAN > 7, the maximum size for shared memory segments is only 512 kB in the default configuration. To display the current value of <TT CLASS="VARNAME" >SHMMAX</TT >, run: </P><PRE CLASS="PROGRAMLISTING" >/etc/conf/bin/idtune -g SHMMAX</PRE ><P> which displays the current, default, minimum, and maximum values. To set a new value for <TT CLASS="VARNAME" >SHMMAX</TT >, run: </P><PRE CLASS="PROGRAMLISTING" >/etc/conf/bin/idtune SHMMAX <TT CLASS="REPLACEABLE" ><I >value</I ></TT ></PRE ><P> where <TT CLASS="REPLACEABLE" ><I >value</I ></TT > is the new value you want to use (in bytes). After setting <TT CLASS="VARNAME" >SHMMAX</TT >, rebuild the kernel: </P><PRE CLASS="PROGRAMLISTING" >/etc/conf/bin/idbuild -B</PRE ><P> and reboot. </P ></DD ></DL ></DIV ><DIV CLASS="TABLE" ><A NAME="SHARED-MEMORY-PARAMETERS" ></A ><P ><B >Table 17-2. <SPAN CLASS="PRODUCTNAME" >PostgreSQL</SPAN > shared memory usage</B ></P ><TABLE BORDER="1" CLASS="CALSTABLE" ><COL><COL><THEAD ><TR ><TH >Usage</TH ><TH >Approximate shared memory bytes required (as of 8.3)</TH ></TR ></THEAD ><TBODY ><TR ><TD >Connections</TD ><TD >(1800 + 270 * <A HREF="runtime-config-locks.html#GUC-MAX-LOCKS-PER-TRANSACTION" >max_locks_per_transaction</A >) * <A HREF="runtime-config-connection.html#GUC-MAX-CONNECTIONS" >max_connections</A ></TD ></TR ><TR ><TD >Autovacuum workers</TD ><TD >(1800 + 270 * <A HREF="runtime-config-locks.html#GUC-MAX-LOCKS-PER-TRANSACTION" >max_locks_per_transaction</A >) * <A HREF="runtime-config-autovacuum.html#GUC-AUTOVACUUM-MAX-WORKERS" >autovacuum_max_workers</A ></TD ></TR ><TR ><TD >Prepared transactions</TD ><TD >(770 + 270 * <A HREF="runtime-config-locks.html#GUC-MAX-LOCKS-PER-TRANSACTION" >max_locks_per_transaction</A >) * <A HREF="runtime-config-resource.html#GUC-MAX-PREPARED-TRANSACTIONS" >max_prepared_transactions</A ></TD ></TR ><TR ><TD >Shared disk buffers</TD ><TD >(<A HREF="runtime-config-preset.html#GUC-BLOCK-SIZE" >block_size</A > + 208) * <A HREF="runtime-config-resource.html#GUC-SHARED-BUFFERS" >shared_buffers</A ></TD ></TR ><TR ><TD >WAL buffers</TD ><TD >(<A HREF="runtime-config-preset.html#GUC-WAL-BLOCK-SIZE" >wal_block_size</A > + 8) * <A HREF="runtime-config-wal.html#GUC-WAL-BUFFERS" >wal_buffers</A ></TD ></TR ><TR ><TD >Fixed space requirements</TD ><TD >770 kB</TD ></TR ></TBODY ></TABLE ></DIV ></DIV ><DIV CLASS="SECT2" ><H2 CLASS="SECT2" ><A NAME="AEN24159" >17.4.2. Resource Limits</A ></H2 ><P > Unix-like operating systems enforce various kinds of resource limits that might interfere with the operation of your <SPAN CLASS="PRODUCTNAME" >PostgreSQL</SPAN > server. Of particular importance are limits on the number of processes per user, the number of open files per process, and the amount of memory available to each process. Each of these have a <SPAN CLASS="QUOTE" >"hard"</SPAN > and a <SPAN CLASS="QUOTE" >"soft"</SPAN > limit. The soft limit is what actually counts but it can be changed by the user up to the hard limit. The hard limit can only be changed by the root user. The system call <CODE CLASS="FUNCTION" >setrlimit</CODE > is responsible for setting these parameters. The shell's built-in command <TT CLASS="COMMAND" >ulimit</TT > (Bourne shells) or <TT CLASS="COMMAND" >limit</TT > (<SPAN CLASS="APPLICATION" >csh</SPAN >) is used to control the resource limits from the command line. On BSD-derived systems the file <TT CLASS="FILENAME" >/etc/login.conf</TT > controls the various resource limits set during login. See the operating system documentation for details. The relevant parameters are <TT CLASS="VARNAME" >maxproc</TT >, <TT CLASS="VARNAME" >openfiles</TT >, and <TT CLASS="VARNAME" >datasize</TT >. For example: </P><PRE CLASS="PROGRAMLISTING" >default:\ ... :datasize-cur=256M:\ :maxproc-cur=256:\ :openfiles-cur=256:\ ...</PRE ><P> (<TT CLASS="LITERAL" >-cur</TT > is the soft limit. Append <TT CLASS="LITERAL" >-max</TT > to set the hard limit.) </P ><P > Kernels can also have system-wide limits on some resources. <P ></P ></P><UL ><LI ><P > On <SPAN CLASS="PRODUCTNAME" >Linux</SPAN > <TT CLASS="FILENAME" >/proc/sys/fs/file-max</TT > determines the maximum number of open files that the kernel will support. It can be changed by writing a different number into the file or by adding an assignment in <TT CLASS="FILENAME" >/etc/sysctl.conf</TT >. The maximum limit of files per process is fixed at the time the kernel is compiled; see <TT CLASS="FILENAME" >/usr/src/linux/Documentation/proc.txt</TT > for more information. </P ></LI ></UL ><P> </P ><P > The <SPAN CLASS="PRODUCTNAME" >PostgreSQL</SPAN > server uses one process per connection so you should provide for at least as many processes as allowed connections, in addition to what you need for the rest of your system. This is usually not a problem but if you run several servers on one machine things might get tight. </P ><P > The factory default limit on open files is often set to <SPAN CLASS="QUOTE" >"socially friendly"</SPAN > values that allow many users to coexist on a machine without using an inappropriate fraction of the system resources. If you run many servers on a machine this is perhaps what you want, but on dedicated servers you might want to raise this limit. </P ><P > On the other side of the coin, some systems allow individual processes to open large numbers of files; if more than a few processes do so then the system-wide limit can easily be exceeded. If you find this happening, and you do not want to alter the system-wide limit, you can set <SPAN CLASS="PRODUCTNAME" >PostgreSQL</SPAN >'s <A HREF="runtime-config-resource.html#GUC-MAX-FILES-PER-PROCESS" >max_files_per_process</A > configuration parameter to limit the consumption of open files. </P ></DIV ><DIV CLASS="SECT2" ><H2 CLASS="SECT2" ><A NAME="AEN24191" >17.4.3. Linux Memory Overcommit</A ></H2 ><P > In Linux 2.4 and later, the default virtual memory behavior is not optimal for <SPAN CLASS="PRODUCTNAME" >PostgreSQL</SPAN >. Because of the way that the kernel implements memory overcommit, the kernel might terminate the <SPAN CLASS="PRODUCTNAME" >PostgreSQL</SPAN > server (the master server process) if the memory demands of another process cause the system to run out of virtual memory. </P ><P > If this happens, you will see a kernel message that looks like this (consult your system documentation and configuration on where to look for such a message): </P><PRE CLASS="PROGRAMLISTING" >Out of Memory: Killed process 12345 (postgres). </PRE ><P> This indicates that the <TT CLASS="FILENAME" >postgres</TT > process has been terminated due to memory pressure. Although existing database connections will continue to function normally, no new connections will be accepted. To recover, <SPAN CLASS="PRODUCTNAME" >PostgreSQL</SPAN > will need to be restarted. </P ><P > One way to avoid this problem is to run <SPAN CLASS="PRODUCTNAME" >PostgreSQL</SPAN > on a machine where you can be sure that other processes will not run the machine out of memory. If memory is tight, increasing the swap space of the operating system can help avoiding the problem, because the out-of-memory (OOM) killer is invoked whenever physical memory and swap space are exhausted. </P ><P > On Linux 2.6 and later, an additional measure is to modify the kernel's behavior so that it will not <SPAN CLASS="QUOTE" >"overcommit"</SPAN > memory. Although this setting will not prevent the <A HREF="http://lwn.net/Articles/104179/" TARGET="_top" >OOM killer</A > from being invoked altogether, it will lower the chances significantly and will therefore lead to more robust system behavior. This is done by selecting strict overcommit mode via <TT CLASS="COMMAND" >sysctl</TT >: </P><PRE CLASS="PROGRAMLISTING" >sysctl -w vm.overcommit_memory=2</PRE ><P> or placing an equivalent entry in <TT CLASS="FILENAME" >/etc/sysctl.conf</TT >. You might also wish to modify the related setting <TT CLASS="LITERAL" >vm.overcommit_ratio</TT >. For details see the kernel documentation file <TT CLASS="FILENAME" >Documentation/vm/overcommit-accounting</TT >. </P ><P > Some vendors' Linux 2.4 kernels are reported to have early versions of the 2.6 overcommit <TT CLASS="COMMAND" >sysctl</TT > parameter. However, setting <TT CLASS="LITERAL" >vm.overcommit_memory</TT > to 2 on a kernel that does not have the relevant code will make things worse not better. It is recommended that you inspect the actual kernel source code (see the function <CODE CLASS="FUNCTION" >vm_enough_memory</CODE > in the file <TT CLASS="FILENAME" >mm/mmap.c</TT >) to verify what is supported in your copy before you try this in a 2.4 installation. The presence of the <TT CLASS="FILENAME" >overcommit-accounting</TT > documentation file should <SPAN CLASS="emphasis" ><I CLASS="EMPHASIS" >not</I ></SPAN > be taken as evidence that the feature is there. If in any doubt, consult a kernel expert or your kernel vendor. </P ></DIV ></DIV ><DIV CLASS="NAVFOOTER" ><HR ALIGN="LEFT" WIDTH="100%"><TABLE SUMMARY="Footer navigation table" WIDTH="100%" BORDER="0" CELLPADDING="0" CELLSPACING="0" ><TR ><TD WIDTH="33%" ALIGN="left" VALIGN="top" ><A HREF="server-start.html" ACCESSKEY="P" >Prev</A ></TD ><TD WIDTH="34%" ALIGN="center" VALIGN="top" ><A HREF="index.html" ACCESSKEY="H" >Home</A ></TD ><TD WIDTH="33%" ALIGN="right" VALIGN="top" ><A HREF="server-shutdown.html" ACCESSKEY="N" >Next</A ></TD ></TR ><TR ><TD WIDTH="33%" ALIGN="left" VALIGN="top" >Starting the Database Server</TD ><TD WIDTH="34%" ALIGN="center" VALIGN="top" ><A HREF="runtime.html" ACCESSKEY="U" >Up</A ></TD ><TD WIDTH="33%" ALIGN="right" VALIGN="top" >Shutting Down the Server</TD ></TR ></TABLE ></DIV ></BODY ></HTML >