RHEL7/CentOS7 check_disk autofs issue

[biox]

Description:
Autofs appears to mount remote autofs disks when it is run, even when the -l flag is specified. Please see this forum post for more information:
https://support.nagios.com/forum/viewtopic.php?f=35&t=37133

Reproduction:
Make an NFS server and use the following config in /etc/exports

/centos7 192.168.x.x(rw,sync,no_subtree_check,no_root_squash)

Connect a server to this NFS server using autofs. My examples is as follows.

mkdir /centos7
yum install nfs-utils autofs -y

grep cent /etc/auto.master
/- /etc/auto.centos7

cat /etc/auto.centos7
/centos7 -rw,soft,intr,rsize=8192,wsize=8192 <nfsserver>:/centos7

systemctl start autofs

umount /centos7

/usr/local/nagios/libexec/check_disk -w 2 -c 1 -l | grep cent
DISK OK - free space: / 16606 MB (92% inode=99%); /dev 478 MB (100% inode=99%); /dev/shm 488 MB (100% inode=99%); /run 482 MB (98% inode=99%); /sys/fs/cgroup 488 MB (100% inode=99%); /boot 332 MB (67% inode=99%);  run/user/0 97 MB (100% inode=99%); /centos7 14756 MB (82% inode=99%);| /=1271MB;17876 17877;0;17878 /dev=0MB;476;477;0;478 /dev/shm=0MB;486;487;0;488 /run=6MB;486;487;0;488 /sys/fs/cgroup=0MB;486;487;0;488 /boot=163MB;494;495;0;496 /run/user/0=0MB;95;96;0;97 /centos7=3161MB;17916;17917;0;17918

You can see that the check_disk command mounts the remote fileshare even though the -l flag is for checking local filesystems only. This works perfectly fine on CentOS 6.

[root@localhost ~]# umount /centos6/
[root@localhost ~]# /usr/local/nagios/libexec/check_disk -w 2 -c 1 -l | grep cent
[root@localhost ~]#

[jfrickson]

I did some looking into this a couple months ago, and didn't find an obvious answer. I'll be looking again soon.

[biox]

Thanks for the update, looking forward to a fix!

[cdawson73]

I was able to get around this by passing a --exclude_device=/etc/auto.r in the check_disk command. This should still be addressed since passing the -l should ignore NFS mounted filesystems. Thanks.

…

On Tue, Mar 21, 2017 at 1:41 PM, Jesse Olson ***@***.***> wrote: Thanks for the update, looking forward to a fix! — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#141 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ACyuMo9Ej1Oi45BQLsbrTL6rTG7mi0CJks5roAu8gaJpZM4HhJ_Z> .

[syl-robitaille]

We're using check_disk in an environment that makes extensive use
of autofs to mount filesystems from remote filers, and have only
recently started using *EL7. Sure enough we've encountered this
problem here, and it causes Nagios to periodically send an alert
if one of the remote filesystems happens to cross the warning or
critical thresholds ... That's not what we want in cases where we've
told check_disk to check only "local" filesystems.

I've dealt with the problem here by patching gl/mountlist.c as follows.
I can't claim to have studied check_disk itself or any other plugins
that might use ME_REMOTE extensively enough to propose that this is
the correct solution in general, but it is rather simple and gives
us exactly the result we're after here, so I decided to submit it in
case it could be useful to others. I hope it helps.

--- ./gl/mountlist.c    2014-03-04 16:41:56.000000000 -0500
+++ ./gl/mountlist.c    2017-04-03 12:43:58.813199509 -0400
@@ -214,7 +214,10 @@
 /* A file system is "remote" if its Fs_name contains a ':'
    or if (it is of type (smbfs or cifs) and its Fs_name starts with '//').  */
+/* 2017-04-04 Sylvain Robitaille (syl@encs.concordia.ca): treat
+   autofs-mounted filesystems as remote */
 # define ME_REMOTE(Fs_name, Fs_type)            \
-    (strchr (Fs_name, ':') != NULL              \
+    (strcmp (Fs_type, "autofs") == 0            \
+     || strchr (Fs_name, ':') != NULL           \
      || ((Fs_name)[0] == '/'                    \
          && (Fs_name)[1] == '/'                 \
          && (strcmp (Fs_type, "smbfs") == 0     \

--
Sylvain Robitaille

[jfrickson]

Fixed in branch maint via commit a29cd35.

Will be in release 2.2.2