The pressure stall information (PSI) metrics provide useful information
on delays caused by waiting for CPU, IO and memory. Particularly on busy
servers it can provide a quick overview of what's "slowing things down".
This feature is supported on Linux >= 4.20.
The interface is documented here:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/plain/Documentation/accounting/psi.txt
These links provide rationale:
https://lwn.net/Articles/759781/https://facebookmicrosites.github.io/psi/
The following metrics are added, corresponding to the currently exposed
lines (see `head /proc/pressure/*`):
- PressureStallCPUSome
- PressureStallIOSome
- PressureStallIOFull
- PressureStallMemorySome
- PressureStallMemoryFull
The color scheme is the same as that used for Load Average, however I
gave it separate entries just in case someone wants to change them
specifically.
Tested on 4.20.7-arch1-1-ARCH, on the linux platform.
Also tested that other platforms still compile (changed configure to use
the unsupported platform).
Closes#879.
Disable the follow process logic in Action_pickFromVector(), when
selecting sort order or user filter, since they don't apply on specific
process.
Fix#856
In the listing of Available Meters for CPUs, the list of CPUs is always
presented by counting them from one. However, if the user prefers to
count CPUs from zero, this is sometimes confusing when fine-tuning the
meters.
Use the appropriate types when calling sysctl().
Currently, `unsigned long long int` is used for all sizes and on
FreeBSD/powerpc this causes all sysctl() calls in scanMemoryInfo()
to fail as they are actually of different sizes on powerpc, where
(sizeof(unsigned long long int), sizeof(u_long)) == (8, 4)
vs (8, 8) on amd64. This results in bogus memory sizes being
reported by htop.
Signed-off-by: Tobias Kortkamp <tobik@FreeBSD.org>
The current OpenBSD-specific CPU usage code is broken. The `cpu`
parameter of `Platform_setCPUValues` is an integer in the interval
[0, cpuCount], not [0, cpuCount-1]: Actual CPUs are numbered from
1, the “zero” CPU is a “virtual” one which represents the average
of actual CPUs (I guess it’s inherited from Linux’s `/proc/stats`).
This off-by-one error leads to random crashes.
Moreover, the displayed CPU usage is more detailed with system,
user and nice times.
I made the OpenBSD CPU code more similar to the Linux CPU code,
removing a few old bits from OpenBSD’s top(1). I think it will be
easier to understand, maintain and evolve.
I’d love some feedback from experienced OpenBSD people.
this way a remount of /proc will not reset starttimes
and we can also see startup times for processes started before the mount
of /proc
also record btime (boot time in seconds since epoch) as Linux semi-global
The source code correctly states that the maximum PID number in
the OpenBSD kernel is fixed in sys/sys/proc.h, however this was
updated in revision 1.215 (two years ago!) from 32766 to 99999.
Introduction of CP_SPIN sched state broke hard-coded state indexes
resulting in the meters incorrectly reporting bogus intr data instead of
CPU usage. Change hardcoded values to sched.h macros.
This would prevent a careless future package maintainer from creating a
release tarball with a defective configure script. :)
Also, add a warning in the autogen.sh phase if pkg.m4 is unused.
Signed-off-by: Kang-Che Sung <explorer09@gmail.com>
With this commit:
* if pkgconfig is installed, it will expand the code inside the shell if, adding the pkgconfig-based tests for the dependencies of Linux delay accounting.
* if pkgconfig is not installed, it will add an error message inside the test of Linux delay accounting, telling the user to install pkgconfig and rebuild the configure script if they want to use Linux delay accounting.
The end result is:
* people running Linux
* will not need pkgconfig when not using delay accounting
* will need pkgconfig when using delay accounting
* if they don't have it
* they are told by configure they need to install it and re-run autogen.sh when running from Git
* they are told by configure they need to install it and re-run configure when running from the tarball
* people not running Linux
* will never need pkgconfig
...and in none of the above scenarios the generated configure script produces unexpanded macros for users checking out the sources from Git.