kata-containers/s390-tools
1
0
Fork 0
mirror of https://github.com/ibm-s390-linux/s390-tools.git synced 2026-08-05 02:14:52 +00:00
Code Issues Packages Projects Releases Wiki Activity
Files
cc6f1a6d7b92aa61806812ab40e7ede8ba288802
s390-tools/cpuplugd/man/cpuplugd.conf.5
Michael Holzheu b627b8d8e1 Initial s390-tools-2.0.0 import 
This commit is based on the s390-tools-1.39.0 version.

Changes on top of s390-tools-1.39.0:

 - Add MIT license to all source files
 - Add LICENSE file
 - Transform REAMDE to README.md (markdown)
 - Add AUTHORS.md file
 - Add CONTRIBUTING.md file
 - Move changelog from README to CHANGELOG.md file

Reviewed-by: Stefan Haberland <sth@linux.vnet.ibm.com>
Signed-off-by: Michael Holzheu <holzheu@linux.vnet.ibm.com>
2017-08-21 10:55:40 +02:00

285 lines
11 KiB
Groff
Raw Blame History

.\" Copyright 2017 IBM Corp.
.\" s390-tools is free software; you can redistribute it and/or modify
.\" it under the terms of the MIT license. See LICENSE for details.
.\"
.TH CPUPLUGD.CONF 5 "May 2011" "s390-tools"
.
.SH NAME
cpuplugd.conf \- Configuration file for the Linux on System z CPU and memory
hotplug daemon
.
.SH DESCRIPTION
The cpuplugd.conf configuration file contains the configuration information for
the Linux for System z CPU and memory hotplug daemon.
Use this file to specify rules for enabling or disabling CPUs and for adding
or removing memory.
.
.SS "CPU hotplug"
CPUs can be enabled and disabled through a sysfs interface.
The status file for a CPU, here CPU number 16 (counting starts at 0),
is /sys/devices/system/cpu/cpu15/online.
Writing a 0 to this file disables the CPU. Writing a 1 enables the CPU.
.
.SS "Memory hotplug"
The rules that add or remove memory use the cooperative memory management
(CMM) feature.
CMM is a mechanism to reduce the memory available to Linux instances that run
as guests of z/VM.
CMM allocates pages to a dynamic page pool not available to Linux.
A diagnose code indicates to z/VM that the pages in the page pool are out of
use. z/VM can then immediately reuse these pages for other guests.
.
.SS "Layout of the configuration file"
The configuration file contains variables specifying static numbers or
expressions. They are of the format \fB<variable>="<value>"\fP and they
need to be specified within one line. Expressions can be specified
to calculate algebraic values or to define boolean rules, which determine
when a hotplug/hotunplug action should be taken. The maximum valid line
length is 2048 characters.
There are case-insensitive pre-defined and case-sensitive user-defined
variables. The configuration file must include specifications for all
pre-defined variables. If a variable is not set, the hotplug function it
applies to, CPU or memory, is disabled.
The only exception to this rule is CMM_DEC, which defaults to the setting
for CMM_INC if omitted. If a pre-defined variable is set more than once, only
the last occurrence is used. User-defined variables must not be set more than
once.
.
.SS "Hotplug rules"
Set these pre-defined variables to an expression that resolves to a boolean
value (true or false). These variables trigger hotplug actions. Setting a
variable to "0" disables the action.
.
.RS 2
.IP "-" 2
\fBHOTPLUG\fP - used to enable CPUs
.IP "-" 2
\fBHOTUNPLUG\fP - used to disable CPUs
.IP "-" 2
\fBMEMPLUG\fP - used to increase the available memory
.IP "-" 2
\fBMEMUNPLUG\fP - used to decrease the amount of memory
.RE
.PP
The following operators can be used in a hotplug rule expression:
.br
.RS 2
.B + * ( ) / - < >
.RE
.br
Furthermore, the boolean operators \fB & \fP (and) \fB|\fP (or) and \fB!\fP
(not) can be used.
If both HOTPLUG and HOTUNPLUG evaluate to true, only the HOTPLUG action is
triggered. If both MEMPLUG and MEMUNPLUG evaluate to true, only the MEMPLUG
action is triggered.
.
.SS "Pre-defined static variables"
The following pre-defined variables can be set only to a static, positive,
numeric value:
.
.RS 2
.IP "-" 2
\fBCPU_MIN\fP - the minimum number of CPUs to keep online (> 0)
.IP "-" 2
\fBCPU_MAX\fP - the maximum number of CPUs to enable (>= 0)
.IP "-" 2
\fBUPDATE\fP - the interval at which cpuplugd evaluates the rules (in seconds,
> 0)
.IP "-" 2
\fBCMM_MIN\fP - the minimum size of the CMM page pool (>= 0)
.IP "-" 2
\fBCMM_MAX\fP - the maximum size of the CMM page pool (>= 0)
.RE
.PP
If the value of CPU_MAX is 0, the overall number of CPUs found in this system
is used as the maximum.
.
.SS "Pre-defined dynamic variables"
The following pre-defined variables can either be set to a static value or to an
algebraic expression:
.
.RS 2
.IP "-" 2
\fBCMM_INC\fP - the amount of pages by which the CMM page pool is increased
if the MEMUNPLUG rule is matched (available system memory is decreased).
.IP "-" 2
\fBCMM_DEC\fP - the amount of pages by which the CMM page pool is decreased
if the MEMPLUG rule is matched (available system memory is increased).
.RE
.PP
The following operators can be used in a dynamic variable expression:
.br
.RS 2
.B + * ( ) / - < >
.RE
.br
.
.SS "User-defined variables"
You can specify complex calculations as user-defined variables, which can then
be used in expressions. User-defined variables are case-sensitive and must not
match a pre-defined variable or keyword. In the configuration file, definitions
for user-defined variables must precede their use in expressions.
Variable names consist of alphanumeric characters (a-z,A-Z,0-9) and
the "_" character, see section \fB"EXAMPLES"\fP for an example (pgscanrate). The
maximum name
length for a variable is 128 characters, and the maximum total size for all
user-defined variables (names + values) is 4096 characters.
.
.SS "Keywords for CPU hotplug rules"
The \fBHOTPLUG\fP and \fBHOTUNPLUG\fP rules can contain the following
pre-defined keywords:
.
.RS 2
.IP "-" 2
\fBloadavg\fP - the current load average
.IP "-" 2
\fBonumcpus\fP - the current number of CPUs which are online
.IP "-" 2
\fBrunnable_proc\fP - the current amount of runnable processes
.IP "-" 2
\fBuser\fP - the current CPU user percentage
.IP "-" 2
\fBnice\fP - the current CPU nice percentage
.IP "-" 2
\fBsystem\fP - the current CPU system percentage
.IP "-" 2
\fBidle\fP - the current CPU idle percentage
.IP "-" 2
\fBiowait\fP - the current CPU iowait percentage
.IP "-" 2
\fBirq\fP - the current CPU irq percentage
.IP "-" 2
\fBsoftirq\fP - the current CPU softirq percentage
.IP "-" 2
\fBsteal\fP - the current CPU steal percentage
.IP "-" 2
\fBguest\fP - the current CPU guest percentage (depends on kernel version: if not reported in /proc/stat, this is set to 0)
.IP "-" 2
\fBguest_nice\fP - the current CPU guest_nice percentage (depends on kernel version: if not reported in /proc/stat, this is set to 0)
.IP "-" 2
\fBcpustat.<name>\fP - data from /proc/stat and /proc/loadavg
.IP "-" 2
\fBtime\fP - floating point timestamp in "seconds.microseconds" since the Unix
Epoch (1970-01-01 00:00:00 +0000 (UTC))
.RE
.PP
The percentage values are accumulated over all online CPUs, so they can vary
between 0 and (100 * \fBonumcpus\fP).
CPU usage data from /proc/stat and /proc/loadavg is accessible by
specifying \fBcpustat.<name>\fP, where \fB<name>\fP can be any of the keywords
described above, plus \fBtotal_ticks\fP. In this case, \fBloadavg\fP,
\fBonumcpus\fP and \fBrunnable_proc\fP
provide the same values as the pre-defined keywords, while the others refer
to the raw timer ticks as reported by /proc/stat, not the percentage.
For example, \fBcpustat.idle\fP reports the timer ticks spent in idle since
system start, and \fBcpustat.total_ticks\fP indicates the sum of all reported
timer ticks, which can be useful for user-defined percentage calculations.
.
.SS "Keywords for memory hotplug rules"
The \fBMEMPLUG\fP and \fBMEMUNPLUG\fP rules can contain the following
pre-defined keywords:
.
.RS 2
.IP "-" 2
\fBapcr\fP - the amount of page cache operations, i.e. pgpin + pgpout from
/proc/vmstat (in 512 byte blocks / second)
.IP "-" 2
\fBfreemem\fP - the amount of free memory (in megabytes)
.IP "-" 2
\fBswaprate\fP - the number of swap operations, i.e. pswpin + pswpout from
/proc/vmstat (in pages / second)
.IP "-" 2
\fBmeminfo.<name>\fP - any value from /proc/meminfo
.IP "-" 2
\fBvmstat.<name>\fP - any value from /proc/vmstat
.IP "-" 2
\fBtime\fP - floating point timestamp in "seconds.microseconds" since the Unix
Epoch (1970-01-01 00:00:00 +0000 (UTC))
.RE
.PP
All values from /proc/meminfo and /proc/vmstat can be used in an expression
by specifying \fBmeminfo.<name>\fP or \fBvmstat.<name>\fP, where \fB<name>\fP
matches a symbol name reported by /proc/meminfo or /proc/vmstat (case
sensitive), e.g. \fBmeminfo.MemTotal\fP.
.
.SS "History function"
There is a history function for the following keywords:
.
.RS 2
.IP "-" 2
\fBcpustat.<name>\fP - data from /proc/stat and /proc/loadavg
.IP "-" 2
\fBmeminfo.<name>\fP - any value from /proc/meminfo
.IP "-" 2
\fBvmstat.<name>\fP - any value from /proc/vmstat
.IP "-" 2
\fBtime\fP - floating point timestamp in "seconds.microseconds" since the Unix
Epoch (1970-01-01 00:00:00 +0000 (UTC))
.RE
.PP
The history levels can be accessed by appending \fB[<history level>]\fP to the
name, where \fB<history level>\fP indicates the amount of past intervals where
the value was gathered. [0] means the current interval (the [0] can be omitted
in this case), [1] means the previous interval, [2] means two intervals ago,
and so on. The history limit is 100. For example, \fBcpustat.system[1]\fP would
indicate the system value from /proc/stat at the previous interval, while
\fBvmstat.pgpgin\fP and \fBvmstat.pgpgin[0]\fP would both mean the current
pgpgin value from /proc/vmstat.
The \fBtime\fP keyword and its history values can be used to calculate values
dependent on time intervals, see section \fB"EXAMPLES"\fP for an example
(pgscanrate).
.
.SH EXAMPLES
A complete configuration file could look like this:
.nf
------------------------------ config file start ------------------------------
UPDATE="5"
CPU_MIN="2"
CPU_MAX="5"
CMM_MIN="0"
CMM_MAX="131072" # 512 MB
pgscan_k="vmstat.pgscan_kswapd_dma + vmstat.pgscan_kswapd_normal + vmstat.pgscan_kswapd_movable"
pgscan_d="vmstat.pgscan_direct_dma + vmstat.pgscan_direct_normal + vmstat.pgscan_direct_movable"
pgscan_k1="vmstat.pgscan_kswapd_dma[1] + vmstat.pgscan_kswapd_normal[1] + vmstat.pgscan_kswapd_movable[1]"
pgscan_d1="vmstat.pgscan_direct_dma[1] + vmstat.pgscan_direct_normal[1] + vmstat.pgscan_direct_movable[1]"
pgscanrate="(pgscan_k + pgscan_d - pgscan_k1 - pgscan_d1) / (time - time[1])"
cache="meminfo.Cached + meminfo.Buffers"
# CMM_INC: 10% of free memory + cache, in 4K pages
CMM_INC="(meminfo.MemFree + cache) / 40"
# CMM_DEC: 10% of total memory in 4K pages
CMM_DEC="meminfo.MemTotal / 40"
HOTPLUG = "(loadavg > onumcpus + 0.75) & (idle < 10.0)"
HOTUNPLUG = "(loadavg < onumcpus - 0.25) | (idle > 50)"
# Plug memory if page scan rate is above 20 pages / sec
MEMPLUG = "pgscanrate > 20"
# Unplug memory while free memory is above 10% of total memory, or cache uses
# more than 50% of total memory
MEMUNPLUG = "(meminfo.MemFree > meminfo.MemTotal / 10) | (cache > meminfo.MemTotal / 2)"
------------------------------ config file end ------------------------------
.fi
The example includes multiple user-defined variables to calculate the page scan
rate with values from /proc/vmstat, as well as the cache size.
\fBAttention:\fP Do not use these example rules on production systems. The
rules have been designed to illustrate the configuration file syntax and are
not suitable for actually governing hotplug actions. Useful rules differ
considerably depending on the workload, resources, and requirements of the
system they are designed for.
.
.SH SEE ALSO
.BR cpuplugd (8)
Reference in New Issue View Git Blame Copy Permalink