Re[2]: [9fans] diskless boot of one terminal off kfs machine

Re[2]: [9fans] diskless boot of one terminal off kfs machine

[steve.simon]

Mmm, the disks are scsi (segate baracuda's on
an Tekram (NCR) controller)

What was strange was the slow speed of the mkfs | mkext.
The machines are 400Mhz PII and a 350Mhz PII connected
by 10Mbps ether so perhaps its just a sign of my ancient kit.

Anyway it works, and I shall be trying it for real as
when I get my venti server and DSL connection going.

Anyone any opinions which scripts in in /bin/services I
can safely copy to /bin/alt.services to run on /alt.net
(the internet) - I want to be able to connect to my
cpu & fs servers over the net but not get hacked (if possible :-)

-Steve

____________________Reply Separator____________________
Subject:    Re: [9fans] diskless boot of one terminal off kfs machine
Author: 9fans@cse.psu.edu
Date:       04/02/03 23:14

> If any of the fossil developers is interested I was using an fs interleaved
> filesystem accross 3 fairly fast disks and got "disk: io=10021 at 8.779ms"
> I got about 25 of these in 3 hours of mkfs | mkext before giving up.
> The times increased slowly about 13ms but the counts stayed at 10010 +- 10.

I think those are normal.  Do you have DMA turned on?
cat /dev/sdC0/ctl and make sure that the dmactl number
is not zero.

Russ

Re: Re[2]: [9fans] diskless boot of one terminal off kfs machine

[Fco.J.Ballesteros]

[-- Attachment #1: Type: text/plain, Size: 324 bytes --]

I'm using fossil on a fs mirror device and speed seems
quite good. My network is 100M though.
Perhaps fs is responsible for your performance problem,
your requests go through yet another user-level file system.
If that's a problem for you, I may do some measures and tuning
to help locate the problem, just let me know.

[-- Attachment #2: Type: message/rfc822, Size: 2751 bytes --]

From: steve.simon@snellwilcox.com
To: 9fans@cse.psu.edu
Subject: Re[2]: [9fans] diskless boot of one terminal off kfs machine
Date: Wed, 05 Feb 2003 10:18:33 +0000
Message-ID: <3658374261@snellwilcox.com>


Mmm, the disks are scsi (segate baracuda's on
an Tekram (NCR) controller)

What was strange was the slow speed of the mkfs | mkext.
The machines are 400Mhz PII and a 350Mhz PII connected
by 10Mbps ether so perhaps its just a sign of my ancient kit.

Anyway it works, and I shall be trying it for real as
when I get my venti server and DSL connection going.

Anyone any opinions which scripts in in /bin/services I
can safely copy to /bin/alt.services to run on /alt.net
(the internet) - I want to be able to connect to my
cpu & fs servers over the net but not get hacked (if possible :-)

-Steve

____________________Reply Separator____________________
Subject:    Re: [9fans] diskless boot of one terminal off kfs machine
Author: 9fans@cse.psu.edu
Date:       04/02/03 23:14

> If any of the fossil developers is interested I was using an fs interleaved
> filesystem accross 3 fairly fast disks and got "disk: io=10021 at 8.779ms"
> I got about 25 of these in 3 hours of mkfs | mkext before giving up.
> The times increased slowly about 13ms but the counts stayed at 10010 +- 10.

I think those are normal.  Do you have DMA turned on?
cat /dev/sdC0/ctl and make sure that the dmactl number
is not zero.

Russ

Re: Re[2]: [9fans] diskless boot of one terminal off kfs machine

[Russ Cox]

> What was strange was the slow speed of the mkfs | mkext.

forsyth pointed out to me that if you run mkfs | mkext with
the source and destination on the same physical disk, it's
likely to be quite slow because kfs and fossil will fight over
the disk, causing a lot of seeking.  it should be much faster to run

	mkfs >/file/on/other/disk
	mkext </file/on/other/disk

in fact, /file/on/other/disk could be a spare raw disk partition,
since the archive has its own eof marker.

Re: Re[2]: [9fans] diskless boot of one terminal off kfs machine

[Fco.J.Ballesteros]

[-- Attachment #1: Type: text/plain, Size: 104 bytes --]

That might be the reason it works fine here, then.
I run just one fossil to service several partitions.

[-- Attachment #2: Type: message/rfc822, Size: 1860 bytes --]

From: "Russ Cox" <rsc@plan9.bell-labs.com>
To: 9fans@cse.psu.edu
Subject: Re: Re[2]: [9fans] diskless boot of one terminal off kfs machine
Date: Wed, 5 Feb 2003 11:04:26 -0500
Message-ID: <a7981f92a6653b12c82dba61c89a5a37@plan9.bell-labs.com>

> What was strange was the slow speed of the mkfs | mkext.

forsyth pointed out to me that if you run mkfs | mkext with
the source and destination on the same physical disk, it's
likely to be quite slow because kfs and fossil will fight over
the disk, causing a lot of seeking.  it should be much faster to run

	mkfs >/file/on/other/disk
	mkext </file/on/other/disk

in fact, /file/on/other/disk could be a spare raw disk partition,
since the archive has its own eof marker.