ERROR: netclass collector failed after 0.001211s: could not get net class info

[brushek]

After running node_exporter:

I get following errors:

Aug 29 23:48:26 CC1-TEST node_exporter[24813]: time="2019-08-29T23:48:26+02:00" level=error msg="ERROR: netclass collector failed after 0.001211s: could not get net class info: error obtaining net class info: failed to read file \"/sys/class/net/bond0/lower_enp2s0f0\": is a directory" source="collector.go:132"

Go version: 1.12.9
node_exporter: 0.18.1

command line to run node_exporter:

node_exporter --web.listen-address 10.20.0.100:9100 --collector.ntp.server 127.0.0.1 --collector.ntp --collector.systemd --collector.netstat.fields="^(.*_(InErrors|InErrs|InDiscards|OutDiscards)|Ip_Forwarding|Ip(6|Ext)_(InOctets|OutOctets)|Icmp6?_(InMsgs|OutMsgs)|TcpExt_(Listen.*|Syncookies.*)|Tcp_(ActiveOpens|PassiveOpens|RetransSegs|CurrEstab)|Udp6?_(InDatagrams|OutDatagrams|NoPorts))$"

[wurenny]

similar error on master branch (each access /metrics):

ERRO[0005] ERROR: netclass collector failed after 0.014723s: could not get net class info: error obtaining net class info: failed to read file "/sys/class/net/br-35c973968f0c/duplex": invalid argument source="collector.go:132"

env info

compile branch:

master

uname -a

Linux yx3312 3.10.0-327.28.3.el7.x86_64 #1 SMP Thu Aug 18 19:05:49 UTC 2016 x86_64 x86_64 x86_64 GNU/Linux

glibc:

glibc-2.17-222.el7.x86_64

analyzed

#ls /sys/class/net/*/duplex | xargs file
/sys/class/net/br1/duplex:             ERROR: cannot read (Invalid argument)
/sys/class/net/br-35c973968f0c/duplex: ERROR: cannot read (Invalid argument)
/sys/class/net/br-6d0e1c7974b6/duplex: ERROR: cannot read (Invalid argument)
/sys/class/net/br-763c26e2c6eb/duplex: ERROR: cannot read (Invalid argument)
/sys/class/net/br-847a59ee600e/duplex: ERROR: cannot read (Invalid argument)
/sys/class/net/br-b1ea6892dc3e/duplex: ERROR: cannot read (Invalid argument)
/sys/class/net/br-bb473a8b9ae8/duplex: ERROR: cannot read (Invalid argument)
/sys/class/net/br-bd6140a9a758/duplex: ERROR: cannot read (Invalid argument)
/sys/class/net/br-f0575229f8f4/duplex: ERROR: cannot read (Invalid argument)
/sys/class/net/docker0/duplex:         ERROR: cannot read (Invalid argument)
/sys/class/net/em1/duplex:             ASCII text
/sys/class/net/em2/duplex:             ERROR: cannot read (Invalid argument)
/sys/class/net/em3/duplex:             ASCII text
/sys/class/net/em4/duplex:             ERROR: cannot read (Invalid argument)
/sys/class/net/lo/duplex:              ERROR: cannot read (Invalid argument)
/sys/class/net/veth11328be/duplex:     ASCII text
/sys/class/net/veth201dc59/duplex:     ASCII text
/sys/class/net/veth45e4727/duplex:     ASCII text
/sys/class/net/veth4a3ed5b/duplex:     ASCII text
/sys/class/net/veth594e7f0/duplex:     ASCII text
/sys/class/net/veth713c958/duplex:     ASCII text
/sys/class/net/veth774fdd0/duplex:     ASCII text

#cat /sys/class/net/br-6d0e1c7974b6/duplex 
cat: /sys/class/net/br-6d0e1c7974b6/duplex: Invalid argument
#
#stat /sys/class/net/br-6d0e1c7974b6/duplex
  File: ‘/sys/class/net/br-6d0e1c7974b6/duplex’
  Size: 4096            Blocks: 0          IO Block: 4096   regular file
Device: 10h/16d Inode: 44374       Links: 1
Access: (0444/-r--r--r--)  Uid: (    0/    root)   Gid: (    0/    root)
Context: system_u:object_r:sysfs_t:s0
Access: 2019-09-03 09:51:19.152032077 +0800
Modify: 2019-09-03 09:51:19.152032077 +0800
Change: 2019-09-03 09:51:19.152032077 +0800
 Birth: -

references:

https://unix.stackexchange.com/questions/319936/cat-write-error-invalid-argument
https://www.kernel.org/doc/Documentation/ABI/testing/sysfs-class-net

[pgier]

This seems to be caused by the recent procfs update and should be fixed by prometheus/procfs#213

[brushek]

After applying this change still get error, but this time it looks like problem is elsewhere:
Sep 11 23:14:26 TEST node_exporter[1590]: time="2019-09-11T23:14:26+02:00" level=error msg="ERROR: netclass collector failed after 0.002598s: could not get net class info: error obtaining net class info: failed to read file \"/sys/class/net/bond0/name_assign_type\": invalid argument" source="collector.go:132"

[root@test /]# file /sys/class/net/bond0/name_assign_type
/sys/class/net/bond0/name_assign_type: ERROR: cannot read `/sys/class/net/bond0/name_assign_type' (Invalid argument)
[root@test /]# cat /sys/class/net/bond0/name_assign_type
cat: /sys/class/net/bond0/name_assign_type: Invalid argument
[root@test /srv/www/common/prometheus/exporters]#

[pgier]

@brushek Yes, I noticed that one also and submitted prometheus/procfs#215
Any chance you can try out that fix and confirm that it resolves the issue?

[brushek]

@pgier, I don't see last commit, and can't use it in go.mod, don't understand why:
https://github.com/prometheus/procfs/tree/master
vs
https://github.com/prometheus/procfs/tree/ebdc8ebb149a43867a5a1e9277de5376c7a53d18

go.mod doesn't allow to use revision version:

go: errors parsing go.mod:
/root/go/src/github.com/prometheus/node_exporter/go.mod:17: invalid module version "ebdc8ebb149a43867a5a1e9277de5376c7a53d18": unknown revision ebdc8ebb149a43867a5a1e9277de5376c7a53d18
Makefile.common:167: recipe for target 'common-lint' failed
make: *** [common-lint] Error 1

[brushek]

@pgier - I did manage to compile it with 'replace' directive in go.mod. Now I don't see any errors. Thank You for patches !

[pgier]

@brushek Thanks for testing! I created a new procfs release with the fix, and PR #1488 for the update.