Hello,

------------------------------
Roberto Stradella
------------------------------

Hi Roberto,

We have a customer who is running their DB2 workload(s) on AWS EC2 instances, using RHEL 8 and Db2 v11.5.7 (with imminent plans to upgrade to RHEL 9 and Db2 12.1).

Their DATA disk setup looks like this:

/dev/nvme3n1 on /data01 type ext4 (rw,relatime,seclabel)
/dev/nvme2n1 on /data02 type ext4 (rw,relatime,seclabel)

So, similar to your XFS setup (only lacking some of your parameters), but I cannot really say how efficient (and optimal) this is...

Hope this helps, if only a little bit...

Regards, Damir

------------------------------
Damir Wilder
Senior Consultant
Triton Consulting
London
------------------------------

We don't use xfs on our DB2,  but I can see how you might want to tune to the underlying RAID or backend I/O structures.

A general search on ideas to use:

Specific mount options recommended in the community and documentation for XFS with Db2 11.5 include: 
noatime and nodiratime: Disables updates to file access time records, reducing metadata writes and improving performance.
inode64: Allows inodes to use 64-bit addresses, which is highly recommended for multi-terabyte filesystems to prevent potential inode space issues.
logbufs=8: Sets the number of in-memory log buffers to the maximum value, which can benefit metadata-intensive workloads.
logbsize=32k (or larger, up to 256k): Sets the size of each log buffer. A larger size (e.g., 32k or 256k) is recommended for file systems with heavy modifications, as is typical in database environments.
nobarrier: Can increase performance by disabling write barriers, but should only be used if the underlying storage hardware (RAID controller with battery-backed cache, or enterprise-grade SSDs) guarantees write integrity.
largeio: Can be used to optimize I/O size, especially if the filesystem was created with specific stripe parameters (swidth).
swalloc: Optimizes allocation for striped volumes, potentially improving performance on RAID configurations. 
Example Mount Entry (/etc/fstab):
/dev/nvmeXnY /db2/data xfs rw,noatime,nodiratime,inode64,logbufs=8,logbsize=32k,nobarrier,largeio,swalloc,nofail 0 0 

 

------------------------------
Greg Sorensen
------------------------------

I tried this out  and these are my resulting options based on my backend RAID I have.  I'll be migrating our data to these and give it a shot.

The "nobarrier" option was invalid or not supported on either my xfs mod  my kernel 4.18.0-553.50.1.el8_10.x86_64

# lsmod | grep xfs
xfs                  1601536  14
libcrc32c              16384  2 nf_tables,xfs

# mkfs.xfs -b size=4096 <LV name>

# mount

<mnt point> type xfs (rw,noatime,nodiratime,seclabel,swalloc,attr2,largeio,inode64,logbufs=8,logbsize=32k,sunit=512,swidth=512,noquota)

/etc/fstab

<LV> /<mount> xfs rw,noatime,nodiratime,inode64,logbufs=8,logbsize=32k,largeio,swalloc,nofail 0 0

------------------------------
Greg Sorensen
------------------------------

Original Message:
Sent: Tue November 18, 2025 09:57 AM
From: Greg Sorensen
Subject: Db2 11.5 Linux RHEL and xfs filesystem

We don't use xfs on our DB2,  but I can see how you might want to tune to the underlying RAID or backend I/O structures.

A general search on ideas to use:

Specific mount options recommended in the community and documentation for XFS with Db2 11.5 include: 
noatime and nodiratime: Disables updates to file access time records, reducing metadata writes and improving performance.
inode64: Allows inodes to use 64-bit addresses, which is highly recommended for multi-terabyte filesystems to prevent potential inode space issues.
logbufs=8: Sets the number of in-memory log buffers to the maximum value, which can benefit metadata-intensive workloads.
logbsize=32k (or larger, up to 256k): Sets the size of each log buffer. A larger size (e.g., 32k or 256k) is recommended for file systems with heavy modifications, as is typical in database environments.
nobarrier: Can increase performance by disabling write barriers, but should only be used if the underlying storage hardware (RAID controller with battery-backed cache, or enterprise-grade SSDs) guarantees write integrity.
largeio: Can be used to optimize I/O size, especially if the filesystem was created with specific stripe parameters (swidth).
swalloc: Optimizes allocation for striped volumes, potentially improving performance on RAID configurations. 
Example Mount Entry (/etc/fstab):
/dev/nvmeXnY /db2/data xfs rw,noatime,nodiratime,inode64,logbufs=8,logbsize=32k,nobarrier,largeio,swalloc,nofail 0 0 

 

------------------------------
Greg Sorensen
------------------------------