From mboxrd@z Thu Jan 1 00:00:00 1970 From: Spam Subject: Re: nvram to protect fs during power loss? Date: Fri, 31 Dec 2004 13:45:32 +0100 Message-ID: <273759260.20041231134532@tnonline.net> References: <1062660450.20041231004008@tnonline.net> <20041231094126.GV1586@hvs.envisage.co.za> Reply-To: Spam Mime-Version: 1.0 Content-Transfer-Encoding: quoted-printable Return-path: list-help: list-unsubscribe: list-post: Errors-To: flx@namesys.com In-Reply-To: <20041231094126.GV1586@hvs.envisage.co.za> List-Id: Content-Type: text/plain; charset="iso-8859-1" To: reiserfs-list@namesys.com Yes, I know that good hardware already do use battery backups and similar. But I was talking about normal consumer hardware; desktops, laptops, etc. I do not know how the internals of reiser4 works, but one of the problems said to be that it cannot be certain about the order and integrity of the data if a power failure occur. But if reiser4 stored some kind of index number, key etc, on regular intervals then that could certainly help in recovering the filesystem to a working state. As it is now you can end up with a severely broken filesystem that doesn't even mount if you have a power failure. IMO a filesystem should do its best to protect the data on it. I thought that perhaps using NVRAM could aid in the recovery after the powerloss, instead of tossing a user "bad error, use --build-sb to fix" during next boot. ~S =20 > On Fri, Dec 31, 2004 at 12:40:08AM +0100, Spam wrote: >>=20 >> There have been some discussions about recovery abilities during >> power loss if write cache is enabled. Some recovery tool I saw once >> used NVRAM to store progress info so that if you had a power loss it >> would be able to resume. > Typically specialized hardware, typically used for NFS (which is stateles= s) doing > "caching" of the writes to enhance performance. >> Perhaps it would be possible for Reiser4 to store some info, like >> time indexes etc in NVRAMwhen it send sync commands to the disk. >> This way it might be possible to avoid corruptions by simply >> verifying (fsck) the data stored after that time index etc? > The way the journals/logs does?? =B4 --=20