[Top] [All Lists]

Re: Atomicity of xfs_fsr -- also isolation?

To: linux-xfs@xxxxxxxxxxx
Subject: Re: Atomicity of xfs_fsr -- also isolation?
From: Florian Weimer <fw@xxxxxxxxxxxxx>
Date: Wed, 16 Feb 2005 20:07:27 +0100
In-reply-to: <20050216132022.GA8898@xxxxxxxxxxxxxxxxxxxxx> (martin f. krafft's message of "Wed, 16 Feb 2005 14:20:22 +0100")
References: <87ll9obz2e.fsf@xxxxxxxxxxxxx> <20050216101917.GA21891@xxxxxxxxxxxxxxxxxxxxx> <87vf8swt4l.fsf@xxxxxxxxxxxxx> <20050216132022.GA8898@xxxxxxxxxxxxxxxxxxxxx>
Sender: linux-xfs-bounce@xxxxxxxxxxx
* martin f. krafft:

> also sprach Florian Weimer <fw@xxxxxxxxxxxxx> [2005.02.16.1407 +0100]:
>> It's possible to have atomic updates without isolation, at least in
>> the database sense of the term (think ACID).  That's why I ask.
> ... not if they introduce conflicts or cause data loss, though.

Non-serializable histories can result in wrong data even if all
transactions involved are atomic.  Of course, this is just a matter of
definitions Yours seems to include isolation. *shrug*

Steve clarified it (thanks!): xfs_fsr provides isolation, but not in
the way I hoped.  I have to stop the application anyway. 8-/ OTOH,
it's better to regenerate the files at the application level, anyway.

<Prev in Thread] Current Thread [Next in Thread>