On 2017-06-01 10:11, Emmanuel DECAEN wrote:
Le 01/06/2017 à 07:21, Sebastiaan Couwenberg a écrit :
notfound 863797 nagios-nrpe/3.0.1-3
thanks
On 05/31/2017 11:05 PM, Emmanuel DECAEN wrote:
And what does nagios-nrpe-server log on the system where the
check_disk
command you claim fails?
May 31 22:46:45 server nrpe[31037]: Running command:
/usr/lib/nagios/plugins/check_disk -w 20% -c 10% -p /var/tmp/mysql
May 31 22:46:45 server nrpe[31037]: Command completed with return
code 2
and output: DISK CRITICAL - /var/tmp/mysql is not accessible: No such
file or directory
May 31 22:46:45 server nrpe[31037]: Return Code: 2, Output: DISK
CRITICAL - /var/tmp/mysql is not accessible: No such file or
directory
I think the problem is related to this "private" mount in
nagios-nrpe-server (extract from /proc/xx/mountinfo):
121 113 8:5
/tmp/systemd-private-b35c254c031041979d3126e02a0c5c51-nagios-nrpe-server.service-MbLbk1/tmp
/var/tmp rw,relatime shared:66 master:28 - ext4 /dev/sda5
rw,data=ordered
The systemd configuration uses PrivateTmp=true, you can override it by
adding a snippet:
/etc/systemd/system/nagios-nrpe-server.service.d/local.conf
With the following content:
[Service]
PrivateTmp=false
See systemd.exec(5) for details about PrivateTmp.
Thanks for the details.
You have a non-standard setup, hence you need to customize the
configuration. This is not a bug in nagios-nrpe, so I'm closing this
issue.
I disagree. Having a configuration using standard check_disk on any
directory can't be considered as "a non-standard setup" (and especially
when it was working correctly in Debian 8)
Your check_disk command is not standard, on Debian systems
/var/tmp/mysql is not a standard directory.
/var/tmp and its subdirectories are either part of the / slice or a
separate /var slice.
Maybe, I'm missing something but there was no "PrivateTmp=true" in
Debian 8 nagios-nrpe-server.
From what I see, using "PrivateTmp=true" in nagios-nrpe-server is a
*change in behavior from Debian 8 to Debian 9*.
NRPE 2.15 in jessie didn't support systemd, NRPE 3.0 in stretch does.
It is a change in behavior, but not a bug.
I think, you can't indicate "not found" on this bug report as it can be
easily reproduced, with a standard check_disk configuration.
It's not a bug, so I removed the affected version.
Don't you think the package upgrade should warn user about this change
when migrating from Debian 8 to Debian 9 ?
No. Standard setups don't have a separate /var/tmp/mysql partition, and
will have /var/tmp as part of the root file system or separate /var.
check_disk via NRPE works as expected on those systems.
You have a non-standard setup where you explicitly want to monitor a
non-standard subdirectory of /var/tmp and so you need to customize your
monitoring to work with your non-standard setup.
On one of my systems I have a separate /tmp and /var and there
check_disk works just fine via NRPE 3.x with PrivateTmp=true.
Kind Regards,
Bas
