On Sat, Oct 11, 2025 at 11:00:27AM -0400, Michael Stone wrote:
On Fri, Oct 10, 2025 at 08:55:38PM +0200, Franco Martelli wrote:
Is it safe and the right thing to do increase the value of:
"stripe_cache_size"? The default value is:
~# cat /sys/block/md0/md/stripe_cache_size
256
no
or play with "speed_limit_max" value, the default is:
~# cat /proc/sys/dev/raid/speed_limit_max
200000
yes, increase this. I usually use a value of 10000000 (effectively "no
limit"). So why does this tunable exist? The general idea is to make
it possible to ensure that a RAID resync doesn't starve out other use
of the array, by setting a lower maximum bandwidth. The value of
200MByte/s was chosen 20 years ago as a number well in excess of drive
speeds...but drives have gotten a lot faster while the default hasn't
changed. A SATA SSD can hit about 550MByte/s on a RAID resync, and
even a good modern hard drive can get over 250MByte/s. NVMe storage
can easily get over a GByte/s, once the speed limit is raised.
Oh, and for RAID5, check /sys/block/md0/md/group_thread_cnt