[pve-devel] [PATCH manager] ui: RRD graphs: display PSI as percent

[pve-devel] [PATCH manager] ui: RRD graphs: display PSI as percent

[Friedrich Weber]

Pressure stall information are actually in percent. This is not
mentioned explicitly in the documentation, but e.g. in the kernel
source [1]:

> The percentage of wall clock time spent in those compound stall
> states gives pressure numbers between 0 and 100 for each resource,
> where the SOME percentage indicates workload slowdowns and the FULL
> percentage indicates reduced CPU utilization:
>
>	%SOME = time(SOME) / period
>	%FULL = time(FULL) / period

Thus, also display them as percent in the GUI.

This reverts commit 087af558631b96ba4fdc40548234e4fa5478b354.

[1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/kernel/sched/psi.c?h=v6.16#n52

Signed-off-by: Friedrich Weber <f.weber@proxmox.com>
---
 www/manager6/node/Summary.js       | 3 +++
 www/manager6/panel/GuestSummary.js | 3 +++
 2 files changed, 6 insertions(+)

diff --git a/www/manager6/node/Summary.js b/www/manager6/node/Summary.js
index c7abd276..9a3f84a4 100644
--- a/www/manager6/node/Summary.js
+++ b/www/manager6/node/Summary.js
@@ -204,6 +204,7 @@ Ext.define('PVE.node.Summary', {
                             fields: ['pressurecpusome'],
                             colors: ['#FFD13E', '#A61120'],
                             store: rrdstore,
+                            unit: 'percent',
                         },
                         {
                             xtype: 'proxmoxRRDChart',
@@ -212,6 +213,7 @@ Ext.define('PVE.node.Summary', {
                             fields: ['pressureiosome', 'pressureiofull'],
                             colors: ['#FFD13E', '#A61120'],
                             store: rrdstore,
+                            unit: 'percent',
                         },
                         {
                             xtype: 'proxmoxRRDChart',
@@ -220,6 +222,7 @@ Ext.define('PVE.node.Summary', {
                             fields: ['pressurememorysome', 'pressurememoryfull'],
                             colors: ['#FFD13E', '#A61120'],
                             store: rrdstore,
+                            unit: 'percent',
                         },
                     ],
                     listeners: {
diff --git a/www/manager6/panel/GuestSummary.js b/www/manager6/panel/GuestSummary.js
index 7780d731..f46006b4 100644
--- a/www/manager6/panel/GuestSummary.js
+++ b/www/manager6/panel/GuestSummary.js
@@ -142,6 +142,7 @@ Ext.define('PVE.guest.Summary', {
                     fields: ['pressurecpusome', 'pressurecpufull'],
                     colors: ['#FFD13E', '#A61120'],
                     store: rrdstore,
+                    unit: 'percent',
                 },
                 {
                     xtype: 'proxmoxRRDChart',
@@ -151,6 +152,7 @@ Ext.define('PVE.guest.Summary', {
                     fields: ['pressureiosome', 'pressureiofull'],
                     colors: ['#FFD13E', '#A61120'],
                     store: rrdstore,
+                    unit: 'percent',
                 },
                 {
                     xtype: 'proxmoxRRDChart',
@@ -160,6 +162,7 @@ Ext.define('PVE.guest.Summary', {
                     fields: ['pressurememorysome', 'pressurememoryfull'],
                     colors: ['#FFD13E', '#A61120'],
                     store: rrdstore,
+                    unit: 'percent',
                 },
             );
         }
-- 
2.47.2



_______________________________________________
pve-devel mailing list
pve-devel@lists.proxmox.com
https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel

[pve-devel] applied: [PATCH manager] ui: RRD graphs: display PSI as percent

[Thomas Lamprecht]

On Thu, 31 Jul 2025 15:51:42 +0200, Friedrich Weber wrote:
> Pressure stall information are actually in percent. This is not
> mentioned explicitly in the documentation, but e.g. in the kernel
> source [1]:
> 
> > The percentage of wall clock time spent in those compound stall
> > states gives pressure numbers between 0 and 100 for each resource,
> > where the SOME percentage indicates workload slowdowns and the FULL
> > percentage indicates reduced CPU utilization:
> >
> >	%SOME = time(SOME) / period
> >	%FULL = time(FULL) / period
> 
> [...]

This was a bit strange for me, I seemingly ingrained some wrong knownledge when
getting first exposed to PSI when that was a new thing back in 2019 or so,
either the docs where worded a bit differently or I just read a bit to much
into the wording (more likely), anyhow thanks for clarifying this, one thing
that I'm still wondering is this then percent for real (i.e. * 100 already) or
a ratio? Or would we need to do *= 100 scaling?

Anyway, as this restores the status quo with rationale it's already an
improvement as is, so: Applied, thanks!

[1/1] ui: RRD graphs: display PSI as percent
      commit: 2753e104a1e6effa4a0ae995f13636a02f3557af


_______________________________________________
pve-devel mailing list
pve-devel@lists.proxmox.com
https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel

Re: [pve-devel] applied: [PATCH manager] ui: RRD graphs: display PSI as percent

[Friedrich Weber]

On 31/07/2025 16:00, Thomas Lamprecht wrote:
> On Thu, 31 Jul 2025 15:51:42 +0200, Friedrich Weber wrote:
>> Pressure stall information are actually in percent. This is not
>> mentioned explicitly in the documentation, but e.g. in the kernel
>> source [1]:
>>
>>> The percentage of wall clock time spent in those compound stall
>>> states gives pressure numbers between 0 and 100 for each resource,
>>> where the SOME percentage indicates workload slowdowns and the FULL
>>> percentage indicates reduced CPU utilization:
>>>
>>> 	%SOME = time(SOME) / period
>>> 	%FULL = time(FULL) / period
>>
>> [...]
> 
> This was a bit strange for me, I seemingly ingrained some wrong knownledge when
> getting first exposed to PSI when that was a new thing back in 2019 or so,
> either the docs where worded a bit differently or I just read a bit to much
> into the wording (more likely), anyhow thanks for clarifying this, one thing
> that I'm still wondering is this then percent for real (i.e. * 100 already) or
> a ratio? Or would we need to do *= 100 scaling?

I think it's fine now -- at least for /proc/pressure/*, the range is
0..100, e.g. when running a fio in parallel:

# cat /proc/pressure/io
some avg10=39.42 avg60=21.00 avg300=5.98 total=49530229
full avg10=39.41 avg60=20.94 avg300=5.96 total=48766096

And this is also correctly displayed in the GUI now.

> 
> Anyway, as this restores the status quo with rationale it's already an
> improvement as is, so: Applied, thanks!

Thanks!

> 
> [1/1] ui: RRD graphs: display PSI as percent
>       commit: 2753e104a1e6effa4a0ae995f13636a02f3557af



_______________________________________________
pve-devel mailing list
pve-devel@lists.proxmox.com
https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel

Re: [pve-devel] applied: [PATCH manager] ui: RRD graphs: display PSI as percent

[Shannon Sterz]

On Thu Jul 31, 2025 at 3:56 PM CEST, Thomas Lamprecht wrote:
> On Thu, 31 Jul 2025 15:51:42 +0200, Friedrich Weber wrote:
>> Pressure stall information are actually in percent. This is not
>> mentioned explicitly in the documentation, but e.g. in the kernel
>> source [1]:
>>
>> > The percentage of wall clock time spent in those compound stall
>> > states gives pressure numbers between 0 and 100 for each resource,
>> > where the SOME percentage indicates workload slowdowns and the FULL
>> > percentage indicates reduced CPU utilization:
>> >
>> >	%SOME = time(SOME) / period
>> >	%FULL = time(FULL) / period
>>
>> [...]
>
> This was a bit strange for me, I seemingly ingrained some wrong knownledge when
> getting first exposed to PSI when that was a new thing back in 2019 or so,
> either the docs where worded a bit differently or I just read a bit to much
> into the wording (more likely), anyhow thanks for clarifying this, one thing
> that I'm still wondering is this then percent for real (i.e. * 100 already) or
> a ratio? Or would we need to do *= 100 scaling?
>
> Anyway, as this restores the status quo with rationale it's already an
> improvement as is, so: Applied, thanks!

if i remember correctly and read the kernel documentation correctly,
this is proper percent (%) [1]. so no additional multiplying necessary.

> The ratios (in %) are tracked as recent trends over ten, sixty, and
> three hundred second windows, which gives insight into short term
> events as well as medium and long term trends.

also did i quick check here on my machine (by compiling pbs):

> ==> /proc/pressure/cpu <==
> some avg10=10.63 avg60=3.66 avg300=0.87 total=54165853

so the `avg10=10.63` should be percent.

[1]: https://www.kernel.org/doc/html/latest/accounting/psi.html

>
> [1/1] ui: RRD graphs: display PSI as percent
>       commit: 2753e104a1e6effa4a0ae995f13636a02f3557af
>
>
> _______________________________________________
> pve-devel mailing list
> pve-devel@lists.proxmox.com
> https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel



_______________________________________________
pve-devel mailing list
pve-devel@lists.proxmox.com
https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel

Re: [pve-devel] applied: [PATCH manager] ui: RRD graphs: display PSI as percent

[Friedrich Weber]

On 31/07/2025 16:08, Shannon Sterz wrote:
> On Thu Jul 31, 2025 at 3:56 PM CEST, Thomas Lamprecht wrote:
>> On Thu, 31 Jul 2025 15:51:42 +0200, Friedrich Weber wrote:
>>> Pressure stall information are actually in percent. This is not
>>> mentioned explicitly in the documentation, but e.g. in the kernel
>>> source [1]:
>>>
>>>> The percentage of wall clock time spent in those compound stall
>>>> states gives pressure numbers between 0 and 100 for each resource,
>>>> where the SOME percentage indicates workload slowdowns and the FULL
>>>> percentage indicates reduced CPU utilization:
>>>>
>>>> 	%SOME = time(SOME) / period
>>>> 	%FULL = time(FULL) / period
>>>
>>> [...]
>>
>> This was a bit strange for me, I seemingly ingrained some wrong knownledge when
>> getting first exposed to PSI when that was a new thing back in 2019 or so,
>> either the docs where worded a bit differently or I just read a bit to much
>> into the wording (more likely), anyhow thanks for clarifying this, one thing
>> that I'm still wondering is this then percent for real (i.e. * 100 already) or
>> a ratio? Or would we need to do *= 100 scaling?
>>
>> Anyway, as this restores the status quo with rationale it's already an
>> improvement as is, so: Applied, thanks!
> 
> if i remember correctly and read the kernel documentation correctly,
> this is proper percent (%) [1]. so no additional multiplying necessary.
> 
>> The ratios (in %) are tracked as recent trends over ten, sixty, and
>> three hundred second windows, which gives insight into short term
>> events as well as medium and long term trends.

Good catch! So the docs *do* mention it, but of course I only grepped
for "percent" ... :)

> 
> also did i quick check here on my machine (by compiling pbs):
> 
>> ==> /proc/pressure/cpu <==
>> some avg10=10.63 avg60=3.66 avg300=0.87 total=54165853
> 
> so the `avg10=10.63` should be percent.
> 
> [1]: https://www.kernel.org/doc/html/latest/accounting/psi.html
> 
>>
>> [1/1] ui: RRD graphs: display PSI as percent
>>       commit: 2753e104a1e6effa4a0ae995f13636a02f3557af
>>
>>
>> _______________________________________________
>> pve-devel mailing list
>> pve-devel@lists.proxmox.com
>> https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel
> 



_______________________________________________
pve-devel mailing list
pve-devel@lists.proxmox.com
https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel