I have posted this once before but haven't really gotten a response from anyone on this question. I am having an issue with hobbit's disk column alarming on nfs mounts with ip's on linux boxes. It look like it reads the ip and sees it as one of the disk usage or available disk. This is what hobbit shows:
red /hd2 (172% used) has reached the PANIC level (95%)
Filesystem Size Used Avail Use% Mounted on
/dev/sda5 20G 1.7G 18G 9% /
/dev/sda1 99M 12M 83M 12% /boot
/dev/sda7 20G 129M 19G 1% /data
/dev/sda8 74G 129M 70G 1% /hd3
/dev/sda6 20G 1.1G 18G 6% /usr/local
172.31.130.29:/ifs/storage1 17T 132G 17T 1% /hd2
Here is my output from the hobbitclients-linux.sh:
echo "[df]"
df -Ph -x none -x tmpfs -x shmfs -x unknown -x iso9660 | sed -e '/^[^ ][^ ]*$/{
N
s/[ ]*\n[ ]*/ /
}'
Now something that I have found that is pretty odd is I have a solaris box that does an nfs mount and it shows it properly in it's disk:
Filesystem size used avail capacity Mounted on /dev/md/dsk/d0 20G 6.5G 13G 34% / swap 9.1G 24K 9.1G 1% /var/run swap 9.1G 272K 9.1G 1% /tmp /dev/md/dsk/d4 24G 9K 24G 1% /data /dev/md/dsk/d3 20G 1.5G 18G 8% /usr/local 172.31.130.29:/ifs/storage1 16T 133G 16T 1% /hd2
Here is my output from the hobbitclients-sunos.sh:
/bin/df -h |egrep -v "proc|fd|mnttab|objfs|ctfs|devices"
Which I changed from this:
echo "[df]"
All of this because Solaris df cannot show multiple fs-types, or exclude
certain fs types.
FSTYPES=/bin/df -n -l|awk '{print $3}'|egrep -v "^proc|^fd|^mntfs|^ctfs|^devfs|^objfs|^nfs"|sort|uniq
if test "$FSTYPES" = ""; then FSTYPES="ufs"; fi
set $FSTYPES
/bin/df -F $1 -k | grep -v " /var/run"
shift
while test "$1" != ""; do
/bin/df -F $1 -k | grep -v " /var/run" | tail +2
shift
done
The reason I did this is to get my size to show in MB instead of Kbytes. But for some reason the solaris box doesn't show the nfs mount with the ip as an alert.
Anyone have any insight on this and a fix being that I am not good with sed and awk's and don't really know if that is my issue or not.
Thanks
Josh
Joshua Krause wrote:
I have posted this once before but haven’t really gotten a response from anyone on this question. I am having an issue with hobbit’s disk column alarming on nfs mounts with ip’s on linux boxes. It look like it reads the ip and sees it as one of the disk usage or available disk.
A simple workaround (although it might be viewed as dodging the question entirely ;) is to exclude NFS file systems from the disk test all together, and just watch the disk usage of the NFS filesystem by monitoring the disk usage on the NFS server itself. This also has the benefit of preventing the hobbit client from hanging during the disk test should the NFS server become unavailable or have issues. Adding a "-l" switch to the df command will accomplish this.
On a somewhat unrelated note, I tried using the -h option for the disk test in my environment and found that it prevented hobbit from correctly sending alerts when filesystems were yellow or red. I'm sure it was a parsing issue, but never took the time to create a patch for do_disk.c to correct it.
Tom
participants (2)
-
forums@triadbiz.rr.com
-
tomg@mcclatchyinteractive.com