Hi
Due to some requirements for storage encryption, we have to move from raw devices to vxfs filesystems for Sybase devices
Storage is SAN based, operating system: RedHat Enterprise Linux with 2.6 kernel, Veritas cluster
Based on whitepaper Managing Workload with ASE (great doc!)
I've been looking into tuning vxfs filesystem and mount options, so far I've got this list:
- fs-aio-max-nr = 1048576
- fs-file-max = 6291456
- increase queue depth from 128 to 1024 in /sys/block/<dev>/queue/nr_requests
- use NOOP scheduler
Enable VxFS kernel option: vx_allow_cloned_naio:
options vxfs vx_allow_cloned_naio=1
Create filesystem:
- align block size with Sybase page size
- enable large files
mount options:
- enable large files
- enable directio
- enable cio
- noatime
- disable journaling (nolog)
What's the impact of using nolog in case of server reboots?
Is this going to cause long fsck times?
What would be good config for option ioerror ?
Let ASE handle any I/O errors?
Any other suggestions?
e.g. pre-allocating files to avoid fragmentation, minimize amount of i-nodes (only few large files)
thanks