[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-Id: <CDF2F321-7726-417D-8163-2909B9DB9B19@gmail.com>
Date: Mon, 13 Nov 2017 07:00:38 -0800
From: Eduard Rozenberg <edrozenberg@...il.com>
To: linux-ext4@...r.kernel.org
Subject: Suggestion: don't default to ext4 lazy init (or warn about it)
My first post to lkml so please go easy on me :).
After mkfs.ext4 on a 20TB volume, the process
ext4lazyinit is now grinding away on the disks
for several hours, and potentially days more
according to other threads I've read. By design
it is using very little I/O (500-1000 K/s) even
though the volume is capable of 200-500 MB/s
and the system is doing nothing ATM.
The constant grinding noise (with HDD's) that
this causes is disruptive in a home or office
environment. Lazy init may also cause undue
worry to users running systems 'at home' -
"could the fs get corrupted if power goes out
while lazy init is going on?"
As I understand:
* There is no way to monitor its % completion
* There is no way to force it to speed up
* Would require a reformat to disable the lazy feature
If I knew I would have disabled lazy init during mkfs
but now I've copied TB's of data to the drive and
would much rather not have to start over.
Would suggest either a clear warning at mkfs ("using
lazy init by default...") or default lazy init to off.
Defaulting to off will cause the least surprise to
most Linux users. For those who have some real
issue with the standard upfront init times they can
easily search for and enable the lazy init feature.
Personally have never had an issue with upfront
init even on 10-30 TB volumes. Enterprises with
PB volumes can do a bit of reading and enable
lazy init even if it defaults to off.
Thank you,
--Ed
Powered by blists - more mailing lists