HyperScale Appliance/Reference Architecture Gluster Volume is reporting incorrect capacity

Issue

• There was a known issue identified with Redhat Gluster where the Capacity is reported incorrectly

• This is due to an incorrect parameter on the Gluster known as shared-brick-count which will count how many bricks share a file system and how the disk space is divided accordingly (within the Gluster)

Methods:

• Method 1 - if you are on the following Service Pack & Hot Fix Pack

  • SP15 & Hot Fix Pack 35

  • SP16 & Hot Fix Pack 27

  • SP17 & Hot Fix Pack 12

• Method 2 - prior to the above Service Pack or Hot Fix Pack

IMPORTANT NOTE - this only needs to be done once and will be applied across the Gluster\, no further remediation is required.

Method 1

• Commvault introduced an additional feature under (/opt/commvault/MediaAgent) which will generate the shared-brick-count script in (/usr/lib64/glusterfs/3.12.2/filter/) for all the nodes

• Navigate to the following location

  • # cd /opt/commvault/MediaAgent

     

• Execute the following script

  • # ./cvavahi.py fix_gluster_volsize

     

• Then execute the following command (only on one node) and this should resolve the size mismatch

  • # gluster volume set (storage pool name) cluster.min-free-inodes 5%

     

• On each of the HyperScale Nodes, log in and verify all the Bricks have "Shared-brick-count" set to "1" using the following command

  • # grep -r "shared-brick-count" /var/lib/glusterd/vols/HyperScale_Pool

     

• Then re-run df -h and confirm that the volume shows the correct size.

Method 2

Do the following on all Nodes

• Navigate to the following location

  • # cd /usr/lib64/glusterfs/3.12.2/filter/

    • Note - You may need to create the 'filter' folder if it doesn't already exist.

    • Use the following command

      • # mkdir filter

         

• Inside the ‘filter’ folder, create a file named “shared-brick-count.sh” and paste the following content to it :

  • # vi shared-brick-count.sh

     

• Input the following content:

#!/bin/bash

sed -i -e 's/option shared-brick-count [0-9]*/option shared-brick-count 1/g' "$1"

• Give execute permissions to the shared-brick-count.sh file on all nodes by executing the command

  • # chmod 755 shared-brick-count.sh

     

Do this on only 1 Node in the Gluster

• Run the following command to force Gluster to update the volume. Run this on only one node.

  • # gluster volume set HyperScale_Pool cluster.min-free-inodes 5%

     

Verify the content

• On each of the HyperScale Nodes login and verify all the Bricks have "Shared-brick-count" set to "1" using the following command

  • # grep -r "shared-brick-count" /var/lib/glusterd/vols/HyperScale_Pool

     

• Then re-run df -h and confirm that the volume shows the correct size.