Bug Fix: SystemCPUfreq fails when any core is offline

[taherkk]

SystemCpufreq function does not check if a core is online before collecting metrics. In the case when a core is offline it fails and does not collect a couple of metrics.

Refer prometheus/node_exporter#2577 for more details

[taherkk]

I have read the readme and steps for contribution and have tried my best to follow it. This my first open source contribution so please guide me if I have missed something.

[taherkk]

Tests are failing because each directory under .../cpu[0-9] has a file named online in Linux except for cpu0. This contains int spicifying if the core is online or not.
I added the file in my local setup manually and the tests are passing.

Please help me in configuring that file using fixtures.

Please review @pgier

[pgier]

This looks good to me in general. I'm just wondering if there is a way to just check /sys/devices/system/cpu/offline which should be empty most of the time (no CPUs offline)?

Also to fix the tests, looks like you'll have to add the relevant online files to the test data.

system_cpu_test.go:103: stat testdata/fixtures/sys/devices/system/cpu/cpu1/online: no such file or directory

[discordianfish]

You also need to sign off your commit, see the failing DCO check above

[taherkk]

Signed off my commit.
Not sure why the lint step is failing.

[discordianfish]

Here is the lint error:

#!/bin/bash -eo pipefail
git diff --exit-code
diff --git a/testdata/fixtures.ttar b/testdata/fixtures.ttar
index cb49947..b80316a 100644
--- a/testdata/fixtures.ttar
+++ b/testdata/fixtures.ttar
@@ -12543,11 +12543,6 @@ Mode: 444
 Directory: fixtures/sys/devices/system/cpu/cpu1
 Mode: 775
 # ttar - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
-Path: fixtures/sys/devices/system/cpu/cpu1/online
-Lines: 1
-1
-Mode: 544
-# ttar - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
 Directory: fixtures/sys/devices/system/cpu/cpu1/cpufreq
 Mode: 775
 # ttar - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
@@ -12606,6 +12601,11 @@ Lines: 1
 <unsupported>
 Mode: 664
 # ttar - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
+Path: fixtures/sys/devices/system/cpu/cpu1/online
+Lines: 1
+1
+Mode: 544
+# ttar - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
 Directory: fixtures/sys/devices/system/cpu/cpu1/thermal_throttle
 Mode: 755
 # ttar - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

Exited with code exit status 1
CircleCI received exit code 1

You probably didn't run make fixtures to create the ttar file

[taherkk]

I'm exploring the idea of checking for offline cpu's in /sys/devices/system/cpu/offline

Sharing the approach here:
Get cpu's from the file.
Expand ranges.
Filter cpu's which not physically present (Though I have a hexacore CPU the offline file shows a range of 12-15 which are not physically present)

[taherkk]

@pgier, I have updated the code according to the approach mentioned earlier.

[discordianfish]

Thanks!

[AdarshdeepCheema]

@taherkk I think it is causing #530. Can u please confirm
FYI @pgier @discordianfish

[taherkk]

raised PR #534 for it @pgier