On 4/15/22 04:47, Chris Hofstaedtler wrote:
Hi Axel, zhenwei,
* Axel Beckert <a...@debian.org> [220414 15:08]:
Since you're asking, I allow myself to cite my reply to your inquiry
with me back then (June 2021):
Thanks!
---------------------------------------------------------------------
Hmmm, do they do the same? Can I test that irqtop from util-linux
somewhere?
Since people seem to expect the irqtop tool from util-linux, I see
multiple options:
1) If the irqtop from util-linux is clearly superior: Drop building
the irqtop package from src:iptables-netflow and let util-linux
generate a transitional package. (Versions should be no problem
with 2.6 < 2.36.)
I more or less built that binary package, because that tool was in
the upstream sources and no such feature was present in Debian so
far and I didn't want to pull in ruby just for a DKMS kernel module.
2) If none of them is clearly superior and they have different feature
sets, using the alternatives system might be an option.
Since I expect both to be just TUI tools without having an API
being used by other tools, different commandline options should not
be an issue here.
3) Rename one of them, e.g. to irqtop-nf or irqtop-ul or so. (Renaming
both of them will be needed for variant 2 anyways.)
In case you intend to add lsirq for bullseye, you could also add
irqtop as irqtop-ul or so (i.e. variant 3b), too. That shouldn't cause
any disturbance IMHO.
---------------------------------------------------------------------
As far as I can see, I didn't get a reply back from you on these
suggestions of mine. Maybe my mail fell through the cracks. But I
think we should take the discussion up again, probably in this bug
report.
Right. I think I forgot to reply back then - sorry.
Experimental should have util-linux-extra 2.38-4+exp1 very soon,
with irqtop installed. Obviously this can only be used for testing.
Personally I think we should have only one irqtop - from my point of
view it does not matter which one. Maybe the new version is
superior. In any case we should not confuse our users.
zhenwei, do you have input on the differences between the existing
irqtop and the new irqtop from util-linux?
Another point which comes to my mind now is that it might make sense
to rename the current irqtop package to irqtop-nf (or irqtop-ruby)
just to make clear that it does not contain the irqtop tool from
util-linux.
Might be an idea. But lets see what the differences are, first.
Thanks,
Chris
Hi, Chris & Axel
The main difference between the two versions:
- original irqtop shows separated interrupt information
- new irqtop shows aggregate interrupt information
Test env: Debian 10; 96 CPUs on a server, 230 characters per line in
termial.
- irqtop (original version) shows uncompleted interrupts(31 / 96 CPUs).
n194-087-081 - irqtop - 2022-04-15 09:42:48 +0800
CPU0 CPU1 CPU2 CPU3 CPU4 CPU5 CPU6 CPU7
CPU8 CPU9 CPU10 CPU11 CPU12 CPU13 CPU14 CPU15 CPU16 CPU17
CPU18 CPU19 CPU20 CPU21 CPU22 CPU23 CPU24 CPU25 CPU26 CPU27
CPU28 CPU29 CPU30 C
cpuUtil: 0.0 0.0 0.4 0.0 1.3 0.0 0.0 0.2
0.0 0.0 0.0 0.0 0.0 0.0 0.2 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.2 0.2 0.2
0.2 0.0 0.2
%irq: 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0
%sirq: 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0
irqTotal: 32 293 477 5 34 7 5 1805
112 51 3 2 28 2 1901 1 13 16
0 6 1 19 2 2 67 29 51 51 42
9 34
i 9: . 2 0 0 0 . . .
. . . . . . . . . .
. . . . . . . . . . .
. .
i 48: . . . . . . 0 .
. . . . . . . . . .
. . . . . . . . . . .
. .
i 49: . . . . . . . .
. . . . . . . . . .
. . . . . . . . . . .
. .
i 50: . . . . . . . .
. . . . . . . . . .
. . . . . . . . . . .
. .
i 51: . . . . . . . .
. . . . . . . . . .
. . . . . . . . . . .
. .
- irqtop (from util-linux) shows aggregate interrupt information.
irqtop | total: 575548749447 delta: 518913 | n148-134-075 | 2022-04-15
10:02:14+08:00
IRQ TOTAL DELTA NAME
LOC 218396041027 393883 Local timer interrupts
RES 217686711923 50039 Rescheduling interrupts
PIN 40532867503 10053 Posted-interrupt notification event
CAL 15012540676 2013 Function call interrupts
PIW 13810059255 57692 Posted-interrupt wakeup event
TLB 8699607720 1597 TLB shootdowns
221 4235495788 88 IR-PCI-MSI 50331656-edge eth0-4
Other enhanced features from the new version:
- sort by several rules, include IRQ, TOTAL, DELTA and NAME.
- specify cpus in list format to monitor.
- specify output columns to print.
- enable/disable per-cpu statistics by specified mode.
- performance improvement. New irqtop written by C uses a little CPU
when running 'irqtop -d 1', the Ruby version spends more time(quite
obvious on a 96 CPUs platform).
--
zhenwei pi
