md and queue write_cache

[Jerome Charaoui]

Hello,

I recently created a raid1 md volume using two Intel DC P3700 NVMe
devices. I noticed that although both NVMe devices report not needing
cache flushes (/sys/block/nvme*/queue/write_cache contains "write
through"), the md volume created by linking them reports the opposite:
/sys/block/md1/queue/write_cache contains "write back".

I'm wondering if this is by design or simply a bug.

My ultimate goal is to use this device to externally place the journal
of an ext4 filesystem sitting on another big, slow, md device.

AIUI, unless md itself requires cache flushes for it own ends, I'll be
able to mount this filesystem with write barriers disabled and safely
gain a significant performance improvement in write speeds.

Thanks,

-- Jerome

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <http://lists.kernelnewbies.org/pipermail/kernelnewbies/attachments/20180525/7de5c761/attachment.sig>