nmon for Linux Activity
Linux performance monitoring on-screen or to CSV file
Brought to you by:
nigelgriffiths
Activity for nmon for Linux
-
nmon for Linux released /nmon16p_32_binaries_feb_2024.tar.gz
-
-
Nigel Griffiths posted a comment on ticket #23
Sorry for neglecting this report. I do not have access to a mainframe and my IBM friends on the team seem to have moved elsewhere. Delighted if you can compile for RHEL and/or SLES - both seem popular on Z. I will upload to the Sourceforge download list and include a "grateful thanks" note. Please confirm the name and/r company reference you would prefer. Cheers, Nigel
-
-
-
-
-
-
-
-
-
-
Stefano modified a comment on ticket #24
Hi Nigel You're right, it's for Linux. I personally use mainly "Gentoo Linux" (secondary distros are Debian, Mint, Arch - in the company I'm working for we use Red Hat), on AMD64 (using both Intel & AMD CPUs).
-
Hi Nigel You're right, it's for Linux. I personally use mainly "Gentoo Linux" (secondary distros are Debian, Mint, Arch), on AMD64 (using both Intel & AMD CPUs).
-
Hi Stefano, Can we clarify this request a little! Is this for the nmon for AIX or the nmon for Linux? I am guessing only Linux as you call the disk "nvme3n1". Simplest flag would be to stop reporting disk stats in KB So the smallest units become MB. What Linux Distro. Distro version and HW (Intel/Power) are you using? Cheer, Nigel
-
Parameter to make values of some metrics stick to a certain "unit" (e.g. KB/MB)?
-
Tom Ficon created ticket #23
NMON 16n support for Linux on Z - s390x
-
-
costinel created ticket #29
dot (.) hotkey doesn't work when displaying disk groups ("g" hotkey when started with -g auto)
-
OS: Ubuntu Linux 18.04LTS
-
-
-
Robert Jacobson posted a comment on ticket #22
was that gdb output helpful? Is there any other way I can help solve this?
-
odd, nmon did not crash when run via gdb. I was however able to recompile with -g (and without -03) and it did still segfault when run normally. Then I ran gdb with the binary and coredump to generate the following output also: I am not mounting/unmounting while nmon is running. # gdb ./nmon_x86_rhel75 coredump GNU gdb (GDB) Red Hat Enterprise Linux 8.2-19.el8 Copyright (C) 2018 Free Software Foundation, Inc. License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html> This is free...
-
odd, nmon did not crash when run via gdb. I was however able to recompile with -g (and without -03) and it did still segfault when run normally. Then I ran gdb with the binary and coredump to generate the following output # gdb ./nmon_x86_rhel75 coredump GNU gdb (GDB) Red Hat Enterprise Linux 8.2-19.el8 Copyright (C) 2018 Free Software Foundation, Inc. License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html> This is free software: you are free to change and redistribute it. There...
-
Hi Robert, Thanks for reporting this. Is it possible to send me the nmon file up to the point it crashed? Also are you mounting and unmounting NFS v4 file-systems while nmon is running? To debug the problem code, we need a stack trace to identify the nmon code calling libc. Assuming you have gdb available and the core file, source code + binary in the current directory. If compiling yourself include -g and may be don't optimize by removing the -O option. $ gdb nmon GNU gdb verion xxxxxxxx GDB Information...
-
example segfault message: Feb 21 13:55:03 myhostname kernel: nmon[3145175]: segfault at 43 ip 00007fa45c8cdfd5 sp 00007ffe6b9d4b98 error 4 in libc-2.28.so[7fa45c801000+1bc000]
-
16n: segfault in libc when writing to file on RHEL8 with NFSv4
-
James Damgar created ticket #28
njmon hits "invalid field format" writing to InfluxDB v2 due to missing fields for "swaps" data
-
Nils Smeds posted a comment on ticket #17
It works for me on a number of different platforms. Would take some effort to find out if it is the compiler or the ncurses library that is the culprit. But it is just a minor annoyance in the top processes output so I have no real issues with it (and I can fix it on the systems I use) These work fine for me: NAME="Ubuntu" VERSION="20.04.5 LTS (Focal Fossa)" gcc (Ubuntu 9.4.0-1ubuntu1~20.04.1) 9.4.0 libncurses-dev/focal,now 6.2-0ubuntu2 amd64 [installed] [nsmeds-x1(nsmeds):nmon] cat /etc/os-release...
-
This works on RHEL with GCC 10 and fails on Ubuntu with GCC 11. As it is largely a cosmetic change, I am going to close the issue. Thanks for reporting and working a fix but there are higher priorities. Thanks Nigel
-
Interesting. Can you share compile line, compiler detailed version and error message. I had no issues compiling it when I submitted it, but I will reinvestigate if needed. The change is trivial so it is a mystery to me why it would fail unless there has been an error in the merge of the patch. On Wed, Dec 7, 2022, 19:00 Nigel Griffiths nigelgriffiths@users.sourceforge.net wrote: This change fails to clean compile of Ubuntu with gcc 11 installed. [patches:#17] https://sourceforge.net/p/nmon/patches/17/...
-
This change fails to clean compile of Ubuntu with gcc 11 installed.
-
LiukeCode posted a comment on ticket #20
Thanks.
-
Can you send me here the output from nfsstat. The manul page nor internet documents the 67 NFS V4 client stat names!
-
nfs4 index access violation in data-collect-mode
-
Planned for release in nmon version 16o
-
Hi, This does not work for my RHEL 8.4 on Power10 hardware with multipath disks. The disk, the multiple paths, partitions and LVM devices all exist as links in /sys/block and their names point to a directory of other files cheers, Nigel From: feature-requests@nmon.p.re.sourceforge.net feature-requests@nmon.p.re.sourceforge.net On Behalf Of costinel Sent: 15 November 2022 11:35 To: [nmon:feature-requests] 12@feature-requests.nmon.p.re.sourceforge.net Subject: [EXTERNAL] [nmon:feature-requests] Re:...
-
Nigel, I think you are worrying too much. Linux kernel exposes the /sys sysfs filesystem since exactly 20 years since about kernel 2.5: https://lwn.net/Articles/14287/ and any distribution has it. The /sys/block path always enumerates all valid block devices. There is simply no running copy of Linux in the entire world that can run the latest nmon, that doesn't have /sys/block Why not give it a try? Say add this simple option and tell me "here is the build, you're on your own, I will silently delete...
-
Hi, If only life was that simple. With multiple Linux Distro and version I focus on: SLES RHEL Ubuntu all have different views on disks / partition and other things that are found in /proc/diskstats. Then there is Debian, Fedora and many more. Then add in different filesystems plus disk technology Mirroring Multipathing Fibre channel disks And disk labels I have been working on a better way to handle that ghastly Linux mess for 20 years. Disk groups and letting the user decide is my approach. If...
-
strace shows this when running iostat: access("/sys/block/sde", F_OK) = 0 access("/sys/block/sde1", F_OK) = -1 ENOENT (No such file or directory) access("/sys/block/sde2", F_OK) = -1 ENOENT (No such file or directory) access("/sys/block/sde3", F_OK) = -1 ENOENT (No such file or directory) access("/sys/block/sde5", F_OK) = -1 ENOENT (No such file or directory) so they only filter the block devices without partitions On Tue, 15 Nov 2022 at 13:31 costinel costinel@users.sourceforge.net wrote: But iostat...
-
But iostat command can identify all block devices - so if they can have a safe sane way to do it, why nmon can not? :-( On Tue, 15 Nov 2022 at 13:28 Nigel Griffiths nigelgriffiths@users.sourceforge.net wrote: Question: Perhaps could you add a command line parameter to hide partitions in normal view and only see whole disk instead? That is exactly the problem - Linux is VERY vague about what is a "partition" and what is a "whole disk". nmon has to cope with dozens disk types, wacko disk names, partitions,...
-
Question: Perhaps could you add a command line parameter to hide partitions in normal view and only see whole disk instead? That is exactly the problem - Linux is VERY vague about what is a "partition" and what is a "whole disk". nmon has to cope with dozens disk types, wacko disk names, partitions, links virtual disks and more. /proc/diskstats is a total mess. The nmon solution is: nmon -g auto Then hit "g". If you do not like nmon's guess of the "whole disks", copy "auto" to "mydiskgroups "and...
-
Use the -n option with njmon. This switches off the outputting of the process id, PID.
-
env: unbutu 22.10. run the njmon, will echo the pid first, Causes script error. add # line:241 if isinstance(eval(line), dict): fixed
-
njmon2influx.py read njmon data error
-
then perhaps could you add a command line parameter to hide partitions in normal view and only see whole disk instead? On Mon, 14 Nov 2022 at 22:55 Nigel Griffiths nigelgriffiths@users.sourceforge.net wrote: status: open --> wont-fix Comment: Diskgroups are there to simplify the disks and increase the usefulness. If diskgroups have multiple disks then it would be confusing to users. [feature-requests:#12] https://sourceforge.net/p/nmon/feature-requests/12/ Allow ". = only busy disks/procs" to work...
-
nimon overwrites http "Host" header
-
Thanks Ruben
-
nmon is not outputting a CPUUTIL000 header line
-
Fixed in nmon16n - actually might have been an earlier release
-
nmonmerge2 doesn´t compile on debian 9.11
-
Also added #include <ctype.h> #include <string.h> Released as nmonmerge3.c
-
-
Where do you see it written as just %CPU and not %%CPU? I don't think I have seen it. If this really is a problem I guess one could keep the %% and change the call to ncurses? From CURSE mvwprintw(padtop, 1, y, "%s", formatstring); To CURSE mvwprintw(padtop, 1, y, formatstring, NULL); Would that work do you think?
-
njmon_collector: ERROR: Parent directory (\) path names not supported character-position
-
Show correct network interfaces data rate in csv file when virtual network interfaces are added/removed dynamically.
-
Fixing these types of issues which also effects disks and filesystems is a massive project and rewrite. njmon is the better tool and does not suffer the problems.
-
Hi Fernando, Interesting idea. nmon does not support dynamic changing of resources like disks, networks, filesystems. It takes too many CPU cycle to straighten the mess out. Your output format will not work with the nmon Analyser, nmonchart or any other nmon graphing tool I am aware of. You could add this sort of function using nmon external data collectors. See https://www.ibm.com/support/pages/nmon-and-external-data-collectors However my new tool njmon or nimon will handle this and let you save...
-
Allow ". = only busy disks/procs" to work for disk groups as well
-
Diskgroups are there to simplify the disks and increase the usefulness. If diskgroups have multiple disks then it would be confusing to users.
-
Support Linux cgroups
-
Francesco, As you state in your own project monitoring cgroups is a complex area. I think I will leave it out of nmon.
-
Panel to monitor running VMs, LXCs, dockers and such
-
Using LABEL to refer to disks.
-
I'm happy to be a guinea pig and help to root the out as they occur :-) On Mon, Nov 14, 2022 at 5:56 AM Nigel Griffiths nigelgriffiths@users.sourceforge.net wrote: status: open --> accepted Priority: 5 --> 2 Comment: This is the Linux curse! As there are differing standards in system file content across distros and even versions within the distro. There is no alternative than to hitting each inconsisancy in turn. [patches:#18] https://sourceforge.net/p/nmon/patches/18/ njmon.linux.v80 Off-by-one...
-
This is a string declared in a local context, used in only one place and as the argument TO a %s directive. There is no printf formatting of the content. Maybe the variable name should be changed to "headertext" instead of current "formatstring" as the variable name suggests it is something it is not used for. The variable goes out of scope immediately after its use in the call to ncurse library. No risk for any side effects if fixing this in my view. On Mon, Nov 14, 2022 at 5:53 AM Nigel Griffiths...
-
TOP stats - reporting owner of processes
-
Make BBBP information optional in csv
-
Suggestion: Be able to select between rolling avg./time interval instead of pr. interval avg.
-
nmon simply does not keep historic data to allow this. Just the current snapshot data and the previous snapshot.
-
Infiniband support - with solution :-)
-
njmon.linux.v80 minor Makefile edits
-
Thanks for reporting this.
-
njmon.linux.v80 Off-by-one os info string copy
-
This is the Linux curse! As there are differing standards in system file content across distros and even versions within the distro. There is no alternative than to hitting each inconsisancy in turn.
-
No more %%CPU in top process displays
-
Hmm! I worry about the behaviour on old compilers, libc and Linux versions. As the %% becomes just % on screen something is parsing the string. Also this is a curses library not standout issue.
-
Wider wide CPU output
-
Thanks for the suggestion and code - I will investigate and try it.
-
I'd really appreciate some feedback on this. I have done the work to support an adaptable number of CPUs for displaying the CPU utilization, so I can submit the next gen of this code that supports up to 1024 CPUs and adapts the number of CPUs displayed to the current screen size.
-
njmon.linux.v80 minor Makefile edits
-
njmon.linux.v80 Off-by-one os info string copy
-
No more %%CPU in top process displays
-
Wider wide CPU output
1 >