Your monitoring system is claiming that your server is running out of free inodes but your disks are not full at all? Maybe you just need to check your filesystem usage and how many files are being used at the same time… Here are some tips to check and identify the inodes being used.
First of all, you can check percentage of inodes used with this simple command:
df -ih
You will see inode usage for your partitions and mouting points:
#dh -ih Filesystem Inodes IUsed IFree IUse% Mounted on rootfs 120K 7.1K 112K 6% / udev 495K 356 495K 1% /dev tmpfs 497K 299 496K 1% /run /dev/mapper/vg-root 120K 7.1K 112K 6% / tmpfs 497K 6 497K 1% /run/lock tmpfs 497K 2 497K 1% /run/shm /dev/sda1 122K 241 122K 1% /boot /dev/mapper/vg-usr 179K 51K 129K 29% /usr /dev/mapper/vg-var 242K 199K 43K 83% /var
Here, we can easily see that /var is using 83% of inodes, which is quite high.
Let’s now identify how those inodes are used with a compound command using find:
find /var -xdev -printf '%h\n' | sort | uniq -c | sort -k 1 -n
You will get a sorted result like this (I truncated to display only the last lines which are the most important):
#find /var -xdev -printf '%h\n' | sort | uniq -c | sort -k 1 -n
52 /var/lib/ucf/cache
106 /var/spool/exim4/input
153 /var/log
184 /var/log/apache2
301 /var/lib/mibs/ietf
319 /var/cache/apt/archives
2289 /var/lib/dpkg/info
10893 /var/spool/postfix/maildrop
We can clearly see that inodes are mostly used by /var/spool/postfix/maildrop. You just have to go in that folder and check if files are useful and/or if you can do some cleaning there.
Then, once files are removed and directory clean, your inodes will be freed and everything will be back to normal!