On Sun, 11 Sep 2005, Christos Zoulas wrote:

> On Sep 12, 12:52am, rpaulo@netbsd.org (rpaulo@netbsd.org) wrote:
> -- Subject: Re: install/31297
>
> | Synopsis: mounting volumes with -o async speeds up file write operations 500% (5 times)
> |
> | State-Changed-From-To: open->feedback
> | State-Changed-By: rpaulo@netbsd.org
> | State-Changed-When: Mon, 12 Sep 2005 00:52:06 +0000
> | State-Changed-Why:
> | Mouting file systems with the async is evil and very dangerous.
> | >From the mount(8) man page:
> |              async       All I/O to the file system should be done asyn-
> |                          chronously.  In the event of a crash, it is
> |                          impossible for the system to verify the integrity of
> |                          data on a file system mounted with this option.  You
> |                          should only use this option if you have an applica-
> |                          tion-specific data recovery mechanism, or are willing
> |                          to recreate the file system from scratch.
> |
> |
> | I think this PR should be closed.
>
> Maybe not for installation... Or at least we should investigate enabling
> softdeps and seeing if that improves performance.

I tried softdeps, with no significant change in time. async dropped the 
time 5 times. I do not think that there is a problem with an aborted 
installation. An aborted installation will probably be bad in many ways.

My discovery came when I saw the disk trash a lot while restoring a 
complete system from tape. I thought that the speed was not reasonable 
and I tried to improve it.

Also using async in single user mode while plugged into an UPS does not 
look dangerous to me.

Peter