Date: Mon, 09 Mar 1998 15:29:05 -0800 (PST) From: Simon Shapiro <shimon@simon-shapiro.org> To: Terry Lambert <tlambert@primenet.com> Cc: lada@ws2301.gud.siemens.co.at, hackers@FreeBSD.ORG, julian@whistle.com, wilko@yedi.iaf.nl, dmlb@ragnet.demon.co.uk, lada@ws2301.gud.siemens.at Subject: Re: SCSI Bus redundancy... Message-ID: <XFMail.980309152905.shimon@simon-shapiro.org> In-Reply-To: <199803092301.QAA04865@usr08.primenet.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 09-Mar-98 Terry Lambert wrote:
...
> This is a salient point. The ISP's who are interested in the HA
> aspects of such servers are more interested because of interruption
> of service issues, more than they are concerned with data-vaulting.
Sure. Clustering correctly, around a resilient and independant disk array
will accomplish both. Each takes out of it what they want (service
non-interruption, data integrity, or both)
> I'm personally more concerned with being able to lock down the gears
> into a known-to-the-OS state, at all times. I can deal with rolling
> incomplete transactions back seperately, if I need transactions.
Please elaborate (your metaphore brings images of broken gears in my lathe
:-)
> The disk write cache is problematic. Most modern disks, when they lose
> DC, do *not* flush the dirty portion of their write cache. Because the
> cache is permitted to reorder operations without regard to their OS
> dependency order, this means that a write cache that's not written
> completely potentially damages dependency ordered data that the OS
> believes has been written.
a. A good controller will allow you to selectively tune the cache, to the
disk's abilities.
b. A good controller will force caches on the attacjed drives to flush
before it ACKs the shutdown command from the O/S.
b. A UPS that will keep the disks running long enough for that. Any
descent disk cabinet/shelf/bay has redundant power supplies, either 2N,
or N+1.
> Dependency ordered data like that created by DOW or Soft Updates
> technologies.
This becomes even more important when considering clustering filesystems.
>From data resilence point of view, it is not as important. See above.
> With disk write caching turned on, I still need a UPS to be able to
> do this reliably, since I have to (1) not add more work to the write
> cache which might potentially push out already delayed writes, and
> (2) cause the disk to flush it's write cache.
A reasonable UPS for a pc, is less than $100.00.
> High availability can also mean "comes back up quickly, and is robust
> in the face of deleterious conditions".
True. To some it means ``I have backup on tape someplace in the drawer'',
to others it means ``I never loose an e-mail message'', while some say ``I
cannot lose service for more than N seconds''. I think we should try and
serve them all.
----------
Sincerely Yours,
Simon Shapiro
Shimon@Simon-Shapiro.ORG Voice: 503.799.2313
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?XFMail.980309152905.shimon>