Telegraf: inputs.disk with mount_points errors in 1.12.x

[boombatower]

Relevant telegraf.conf:

[[inputs.disk]]
  mount_points = [
    "/hostfs/vol/ssd/0_400G",
    "/hostfs/vol/hdd/0_3T",
    "/hostfs/vol/hdd/2_1T",
    "/hostfs/vol/hdd/3_4T",
    "/hostfs/vol/hdd/4_4T",
  ]

System info:

• works: telegraf 1.11.5 (1.11.x all work) via official docker
• fails: telegraf 1.12.1 (1.12.0 also fails) via official docker

Steps to reproduce:

1. start telegraf with above config

Expected behavior:

works as before

Actual behavior:

2019-09-12T19:21:40Z E! [inputs.disk] Error in plugin: error getting disk usage info: lstat /dev/mapper/vol_ssd_0_400G_crypt: no such file or directory

[danielnelson]

How are you starting the container, are there any environment variables set?

From inside the docker container, on both 1.11.5 and 1.12.1, run:

cat /proc/self/mountinfo
ls -l /dev/mapper/

[boombatower]

...
  --privileged \
  -v /srv/telegraf:/etc/telegraf:ro \
  -v /var/run/docker.sock:/var/run/docker.sock \
  -v /:/hostfs:ro \
  -e HOST_MOUNT_PREFIX=/hostfs \

• mountinfo.txt
• dev_mapper.txt

As expected nothing in dev_mapper in either version as the mount setup is identical, just telegraf is now expecting more.

[danielnelson]

I believe this is fixed upstream by https://github.com/shirou/gopsutil/commit/93e08ea089e8ede9d53a3d345ea4384b08e86ee2.

If I get you a telegraf binary with the fix included would you be able to test it? Probably will want to just mount it over the telegraf in the container.

[danielnelson]

Here is a build for testing the fix (ignore the wrong version number, this is the latest from the 1.12.x branch):

• telegraf-1.12.0~e7208f25_linux_amd64.tar.gz

[boombatower]

If I get you a telegraf binary with the fix included would you be able to test it?

Yes, I can do that.

[boombatower]

I copied my config and only ran disk and output to stdout. I mounted the binaries from /usr/bin of tarball via: -v /root/telegraf:/usr/bin/telegraf:ro and ran latest (1.12.1) docker image.

With the binary override mount proper output was generated:

disk,device=mapper/vol_hdd_2_1T_crypt,fstype=ext4,mode=ro,path=/vol/hdd/2_1T used=18886762496i,used_percent=2.0234354188704478,inodes_total=61054976i,inodes_free=61010063i,inodes_used=44913i,total=983427665920i,free=914514042880i 1568346480000000000

Without the binary override mount it failed like normal:

[inputs.disk] Error in plugin: error getting disk usage info: lstat /dev/mapper: no such file or directory

As such I can confirm your build works.

[xkfen]

I copied my config and only ran disk and output to stdout. I mounted the binaries from /usr/bin of tarball via: -v /root/telegraf:/usr/bin/telegraf:ro and ran latest (1.12.1) docker image.

With the binary override mount proper output was generated:

disk,device=mapper/vol_hdd_2_1T_crypt,fstype=ext4,mode=ro,path=/vol/hdd/2_1T used=18886762496i,used_percent=2.0234354188704478,inodes_total=61054976i,inodes_free=61010063i,inodes_used=44913i,total=983427665920i,free=914514042880i 1568346480000000000

Without the binary override mount it failed like normal:

[inputs.disk] Error in plugin: error getting disk usage info: lstat /dev/mapper: no such file or directory

As such I can confirm your build works.

hi, i had the same problem:
[inputs.disk] Error in plugin: error getting disk usage info: lstat /dev/mapper: no such file or directory.
how did you solve this problem? thanks

[andyshutak]

Yea im also having that problem on the latest docker container

[cookandy]

Same! Can we get a new docker release with the fix please?

[glinton]

I believe this will be going out in the release on Tuesday. If you are unable to wait that long,
consider using a nightly build which should also work.

[danielnelson]

Closed in #6389