Linux + what are the preferred options in order to resolve inodes that reached 100%

Question

we have RHEL machine , and from df -i , we can see that some partitions are with 100% ( about inodes ) , in spite by df -h we have space

Note - disks are VMDK disks

 df -h
/dev/sdc                     40G   17G   23G  43% /data/sdc
/dev/sdd                     40G   23G   17G  58% /data/sdd
/dev/sde                     40G   23G   17G  58% /data/sde
/dev/sdb                     40G   26G   14G  65% /data/sdb


 df -i 
/dev/sdc                    2621440  231948  2389492    9% /data/sdc
/dev/sdd                    2621440 2616820     4620  100% /data/sdd
/dev/sde                    2621440 2613218     8222  100% /data/sde
/dev/sdb                    2621440 2621440        0  100% /data/sdb

so I just collect some options in order to solve the inodes that reached 100% problem

1. Extended the disks from 40G to 100G

then , rescan the disks on OS

 echo 1 >/sys/block/${disk_name}/device/rescan

then resize the disk as

resize2fs /dev/$disk_name

2. the same step as option 1 but additionally create new filesystem as mkfs.ext4 -j -m 0 /dev/$disk -F , in order to increase the inodes by mkfs according to the new disk space

so according to step1 and step 2

is it enough to do only step 1 or also step 2 additionally to step 1?

Answer 1

If you follow point 1 - extend the disks and the extend the filesystem you will be fine.

But rescan may no work as you expect if the disks are busy and you will need to reboot the machine.

About formatting the disks (with recognized new size) - yes, this is also possible. You should decide for self which one is more convenient: wait for filesystems extend or wait for replication of HDFS from other replicas.