Re: [9fans] fs hardware configuration

Re: [9fans] fs hardware configuration

[Fco.J.Ballesteros]

>> Yes, they're. Our local fs is now the one distributed.
>> We recompiled and booted the resulting binary just to be sure.
>
> With or without the "allow/disallow" commands?  I disabled the former,
> left the latter because I feel more comfortable that way.  But what is
> the released standard?

With them in place.

Re: [9fans] fs hardware configuration

[Lucio De Re]

On Mon, Dec 02, 2002 at 03:27:51PM +0100, Fco.J.Ballesteros wrote:
> >
> > With or without the "allow/disallow" commands?  I disabled the former,
> > left the latter because I feel more comfortable that way.  But what is
> > the released standard?
>
> With them in place.

I would recommend re-enabling the password prompt by default, then.
It's better than a totally exposed file server.

Nemo,

I seem to have lost my carefully concocted /sys/src/fs source tree.
You had a rather distinguished Spanish name for your server, could
you jolt my memory, help me find the lost details?

Thanks.  Lucio.

Re: [9fans] fs hardware configuration

[Fco.J.Ballesteros]

When I wrote:
> Our local fs is now the one distributed

I tried to say just that what we're running now is the file server
code as distributed (with no further changes). Since it now includes
the changes to support our disks, we have no reason to run a different one.

Excuse me in case what I wrote could also mean something
different.

Re: [9fans] fs hardware configuration

[Fco.J.Ballesteros]

[-- Attachment #1: Type: text/plain, Size: 414 bytes --]

Well, for me, allow is just to too dangerous and we use it just
to upgrade system files including wstats. Before using it you can
always ask who to the file server. And it's sensible to put an
'echo dont forget to disllow' in whatever scripts you use to install
files. If you fell you really need it, I might change that for you
when I get some time, but I'd prefer not to put that into our regular
kernel.

[-- Attachment #2: Type: message/rfc822, Size: 1628 bytes --]

From: Kenji Arisawa <arisawa@ar.aichi-u.ac.jp>
To: 9fans@cse.psu.edu
Subject: Re: [9fans] fs hardware configuration
Date: Tue, 3 Dec 2002 08:53:26 +0900
Message-ID: <400A7402-0651-11D7-9E52-000393A941BC@ar.aichi-u.ac.jp>

Hello,

 >With or without the "allow/disallow" commands?  I disabled the former,
 >left the latter because I feel more comfortable that way.
I might asking too much...
I would like to have allow command that allows limited users, i.e.,
allow bootes arisawa

Kenji Arisawa

Re: [9fans] fs hardware configuration

[Kenji Arisawa]

> Well, for me, allow is just to too dangerous and we use it just
> to upgrade system files including wstats. Before using it you can
> always ask who to the file server. And it's sensible to put an
> 'echo dont forget to disllow' in whatever scripts you use to install
> files. If you fell you really need it, I might change that for you
> when I get some time, but I'd prefer not to put that into our regular
> kernel.
>
Thank you very much!
I asked:
	allow bootes arisawa
however a single user will be enough for the argument of allow.

Re: [9fans] fs hardware configuration

[Geoff Collyer]

Strictly speaking, it's my local fs kernel (which includes code from
forsyth via nigel and IDE support derived from nemo's) that's now
distributed.  I updated fs(8) and fsconfig(8), so you can see what's
new.

Re: [9fans] fs hardware configuration

[Fco.J.Ballesteros]

[-- Attachment #1: Type: text/plain, Size: 395 bytes --]

> I would recommend re-enabling the password prompt by default, then.
> It's better than a totally exposed file server.

Our fs is at my office, which I consider secure.

> I seem to have lost my carefully concocted /sys/src/fs source tree.
> You had a rather distinguished Spanish name for your server, could

Our fs is called "sargazos".
Drop me a line if I can help somehow.

hth

[-- Attachment #2: Type: message/rfc822, Size: 2364 bytes --]

From: Lucio De Re <lucio@proxima.alt.za>
To: 9fans@cse.psu.edu
Subject: Re: [9fans] fs hardware configuration
Date: Mon, 2 Dec 2002 16:40:18 +0200
Message-ID: <20021202164018.T22558@cackle.proxima.alt.za>

On Mon, Dec 02, 2002 at 03:27:51PM +0100, Fco.J.Ballesteros wrote:
> >
> > With or without the "allow/disallow" commands?  I disabled the former,
> > left the latter because I feel more comfortable that way.  But what is
> > the released standard?
>
> With them in place.

I would recommend re-enabling the password prompt by default, then.
It's better than a totally exposed file server.

Nemo,

I seem to have lost my carefully concocted /sys/src/fs source tree.
You had a rather distinguished Spanish name for your server, could
you jolt my memory, help me find the lost details?

Thanks.  Lucio.

Re: [9fans] fs hardware configuration

[Fco.J.Ballesteros]

> Also, are the idefs changes (dorman/geoff/nemo) in the latest
> distribution?

Yes, they're. Our local fs is now the one distributed.
We recompiled and booted the resulting binary just to be sure.

Re: [9fans] fs hardware configuration

[Lucio De Re]

On Mon, Dec 02, 2002 at 09:31:57AM +0100, Fco.J.Ballesteros wrote:
>
> > Also, are the idefs changes (dorman/geoff/nemo) in the latest
> > distribution?
>
> Yes, they're. Our local fs is now the one distributed.
> We recompiled and booted the resulting binary just to be sure.

With or without the "allow/disallow" commands?  I disabled the former,
left the latter because I feel more comfortable that way.  But what is
the released standard?

++L

Re: [9fans] fs hardware configuration

[Kenji Arisawa]

Hello,

 >With or without the "allow/disallow" commands?  I disabled the former,
 >left the latter because I feel more comfortable that way.
I might asking too much...
I would like to have allow command that allows limited users, i.e.,
allow bootes arisawa

Kenji Arisawa

RE: [9fans] fs hardware configuration

[Skip Tavakkolian]

> If those IDE drivers are IBM 120GXPs, have you checked
> the warranty doesn't preclude 24 7 working?

They are WD.  I had read disturbing stats about Fujitsu drives
failure rates but not IBM's.  The Fujitsu story is here:

http://www.theregister.co.uk/content/63/27926.html

RE: [9fans] fs hardware configuration

[Nigel Roles]

For IBM 120GXPs:

http://www.sudhian.com/docs.cfm/id/97.sud

-----Original Message-----
From: 9fans-admin@cse.psu.edu [mailto:9fans-admin@cse.psu.edu]On Behalf
Of Skip Tavakkolian
Sent: 30 November 2002 13:42
To: 9fans@cse.psu.edu
Subject: RE: [9fans] fs hardware configuration


> If those IDE drivers are IBM 120GXPs, have you checked
> the warranty doesn't preclude 24 7 working?

They are WD.  I had read disturbing stats about Fujitsu drives
failure rates but not IBM's.  The Fujitsu story is here:

http://www.theregister.co.uk/content/63/27926.html

Re: [9fans] fs hardware configuration

[William Josephson]

On Sat, Nov 30, 2002 at 08:42:11AM -0500, Skip Tavakkolian wrote:
> > If those IDE drivers are IBM 120GXPs, have you checked
> > the warranty doesn't preclude 24 7 working?
>
> They are WD.  I had read disturbing stats about Fujitsu drives
> failure rates but not IBM's.  The Fujitsu story is here:

All the manufacturers have had their problems with particular
models.  The 120GXPs (I think) from IBM definitely had major
problems and Fujitsu has had trouble, too.  Western Digital
seems to be known in the industry for very poor QA; it has
been a long time since I've used WD drives, but I did have
quite a bit of trouble when I last used them.  I'm not sure
what to conclude other than "be paranoid if you care about
what is on the disk" :-/

Re: [9fans] fs hardware configuration

[Skip Tavakkolian]

Thanks much. It is exactly what I was looking for.

Re: [9fans] fs hardware configuration

[Geoff Collyer]

The latest file server code and manual pages on sources include IDE
support, the mirror device and a few other small changes.  So the
latest CD image should include them too.  The new fsconfig(8) includes
a more complex real-life example than before.

My file server with 4 IDE disks uses this configuration:

	config h0
	service fs
	filsys main c{p(h0)0.25p(h2)0.25}f{p(h0)25.75p(h2)25.75}
	filsys dump o
	filsys spare h1
	filsys other h3
	ipauth 0.0.0.0
	ipsntp 10.9.0.3
	ip0 10.9.0.1
	ipgw0 10.9.0.254
	ipmask0 255.255.0.0
	end

It's evolved over time and so isn't necessarily the best use of the
drives.  main uses the first 25% of h0 as cache and mirrors it to the
first 25% of h2 (h1.0.0).  It also uses the rest of h0 as a fake worm
and mirrors it to the rest of h2.

For your hardware, I'd suggest a fake worm on one IDE drive, mirrored
to the second, with one SCSI drive as "other" and the other as cache.
So something like this:

	config h0
	service fs
	filsys main cw0f{h0h2}
	filsys dump o
	filsys other w1
	ipauth 0.0.0.0
	ipsntp 10.9.0.3
	ip0 10.9.0.1
	ipgw0 10.9.0.254
	ipmask0 255.255.0.0
	ream main
	ream other
	end

This assumes that you make the IDE drives masters on separate
controllers.  Of course, you only ream the file systems once, to
initialise them.

[9fans] fs hardware configuration

[Skip Tavakkolian]

I'm finally going to build a proper fileserver.  I have a PC with
512MB of memory, symbios scsi controller and 2×9GB SCSI drives.  I'm
also getting 2×120GB IDE drives.  My limited understand leads me to
believe that I could designate the IDE drives as WORM and the SCSI
drives+memory as cache.  What is the best way to configure this?

Also, are the idefs changes (dorman/geoff/nemo) in the latest
distribution?

RE: [9fans] fs hardware configuration

[Nigel Roles]

If those IDE drivers are IBM 120GXPs, have you checked
the warranty doesn't preclude 24 7 working?

-----Original Message-----
From: 9fans-admin@cse.psu.edu [mailto:9fans-admin@cse.psu.edu]On Behalf
Of Skip Tavakkolian
Sent: 29 November 2002 15:54
To: 9fans@cse.psu.edu
Subject: [9fans] fs hardware configuration


I'm finally going to build a proper fileserver.  I have a PC with
512MB of memory, symbios scsi controller and 2×9GB SCSI drives.  I'm
also getting 2×120GB IDE drives.  My limited understand leads me to
believe that I could designate the IDE drives as WORM and the SCSI
drives+memory as cache.  What is the best way to configure this?

Also, are the idefs changes (dorman/geoff/nemo) in the latest
distribution?