Printable View
Hi,
does raidframe work under sparc64? If so how do I configure the kernel to
use it? I've got sys.tar.gz and looked in sparc64/conf/ and in GENERIC
there is nothing about RAIDframe. But under i386/conf/GENERIC there is.
Does this mean it doesn't work under sparc64? I can see a ref to the raid
frame driver in one of the .c files.
Anyone used RAIDframe on sparc64? Is it any good? I've head it can lock up
every 22 hours?
*************************************
This email has been sent to you.
On Mon, 19 Nov 2007 12:11:49 +0000, Khalid Schofield wrote:
[color=blue]
> Hi,
> does raidframe work under sparc64?[/color]
Google with three keywords: openbsd raidframe sparc64. See what comes
up. If that's not enough, there is a raid(4) man page for sparc64:
[url]http://www.openbsd.org/cgi-bin/man.cgi?query=raid&apropos=0&sektion=0&manpath=OpenBSD+Current&arch=sparc64&format=html[/url]
[color=blue]
> If so how do I configure the kernel to
> use it? I've got sys.tar.gz and looked in sparc64/conf/ and in GENERIC
> there is nothing about RAIDframe. But under i386/conf/GENERIC there is.
> Does this mean it doesn't work under sparc64? I can see a ref to the raid
> frame driver in one of the .c files.[/color]
I've not used it on sparc64, but as it is not an architecture dependent,
you can certainly configure it into a kernel and test it yourself.
[color=blue]
> Anyone used RAIDframe on sparc64? Is it any good? I've head it can
> lock up every 22 hours?[/color]
22 hours? Must be due to the sparc64's wind-up mouse powering down. :)
Sheesh! Cite a source for that, if you want it treated seriously.
I've been running RAIDframe on i386 with root-on-raid for several years,
and trust it completely. Have I had problems? Sure. Have other people?
You bet. But I have never lost data, and it has shut down bad hard drives
and kept on running for me.
--
Replying directly will get you locally blacklisted. Change the address;
use my first name in front of the @ if you want to communicate privately.
On Mon, 19 Nov 2007, Josh Grosse wrote:
[color=blue]
> On Mon, 19 Nov 2007 12:11:49 +0000, Khalid Schofield wrote:
>[color=green]
>> Hi,
>> does raidframe work under sparc64?[/color]
>
> Google with three keywords: openbsd raidframe sparc64. See what comes
> up. If that's not enough, there is a raid(4) man page for sparc64:
>
> [url]http://www.openbsd.org/cgi-bin/man.cgi?query=raid&apropos=0&sektion=0&manpath=OpenBSD+Current&arch=sparc64&format=html[/url]
>[/color]
did that at first. Found vague refs to netbsd and openbsd but nothing
recent.
[color=blue][color=green]
>> If so how do I configure the kernel to
>> use it? I've got sys.tar.gz and looked in sparc64/conf/ and in GENERIC
>> there is nothing about RAIDframe. But under i386/conf/GENERIC there is.
>> Does this mean it doesn't work under sparc64? I can see a ref to the raid
>> frame driver in one of the .c files.[/color]
>
> I've not used it on sparc64, but as it is not an architecture dependent,
> you can certainly configure it into a kernel and test it yourself.
>[/color]
compiled the kernel but it hangs and says the normal boot messages up
until it gets to:
Kernelized RAIDframe activated then grinds to a halt and says nothing
more.
[color=blue][color=green]
>> Anyone used RAIDframe on sparc64? Is it any good? I've head it can
>> lock up every 22 hours?[/color]
>
> 22 hours? Must be due to the sparc64's wind-up mouse powering down. :)
> Sheesh! Cite a source for that, if you want it treated seriously.
>[/color]
may be a random complaint that has little technical backing
[color=blue]
> I've been running RAIDframe on i386 with root-on-raid for several years,
> and trust it completely. Have I had problems? Sure. Have other people?
> You bet. But I have never lost data, and it has shut down bad hard drives
> and kept on running for me.
>[/color]
What kind of problems? What error reporting do you get when raidframe
detects bad disks?
[color=blue]
> --
> Replying directly will get you locally blacklisted. Change the address;
> use my first name in front of the @ if you want to communicate privately.
>
>[/color]
On Tue, 20 Nov 2007 14:35:42 +0000, Khalid Schofield wrote:
[color=blue]
> compiled the kernel but it hangs and says the normal boot messages up
> until it gets to:
>
> Kernelized RAIDframe activated then grinds to a halt and says nothing
> more.[/color]
Do you have RAID_AUTOCONFIG in the kernel? If so, you might first try
without it.
You don't say what you're running; there was a problem for some
RAIDframe users on -current between October 10 and November 2 that had
that very symptom.
[color=blue]
> What kind of problems? What error reporting do you get when raidframe
> detects bad disks?[/color]
Two of my problems have come from running -current.
One of the problems I had was a kernel hang on boot as you've just
experienced yourself. It occurred after art@'s context switch changes to
-current, as described above and also here:
[url]http://marc.info/?l=openbsd-misc&m=119289255002826&w=2[/url]
Here's another, caused by disklabel changes for FFS2:
[url]http://marc.info/?l=openbsd-misc&m=118158728932321&w=2[/url]
The only other problem was misunderstanding the limits of
root-on-raid:
[url]http://marc.info/?l=openbsd-misc&m=115668370718537&w=2[/url]
When RAIDframe fails a disk, it produces a kernel message. I have a cron
job that checks raidctl -s output, and reports discrepancies. I've had
two failures caught by RAIDframe, and both times, was alerted by e-mail.
I've also had a disk problem NOT caught by RAIDframe... caught by
smartmontools instead. But that is perfectly understandable, as
smartmontools caught the error during a long offline test where every
sector is checked. RAIDframe will only report errors on I/O ops it
makes.
[color=blue]
>[color=green]
>> --
>> Replying directly will get you locally blacklisted. Change the address;
>> use my first name in front of the @ if you want to communicate
>> privately.
>>
>>[/color][/color]
--
Replying directly will get you locally blacklisted.
Change the address; use my first name in front of the @ if you want to
communicate privately.
[color=blue]
> On Tue, 20 Nov 2007 14:35:42 +0000, Khalid Schofield wrote:
>[color=green]
>> compiled the kernel but it hangs and says the normal boot messages up
>> until it gets to:
>>
>> Kernelized RAIDframe activated then grinds to a halt and says nothing
>> more.[/color]
>
> Do you have RAID_AUTOCONFIG in the kernel? If so, you might first try
> without it.[/color]
Yup have RAID_AUTOCONFIG in the kernel. So I was following an example on
setting up a raid 1 two disk mirror on a web page. It said make wd0d and
wd1d which will have the OS on it and wd0a and wd1a will have a basic
system installed so the system will boot. I marked wd0d and wd1d as RAID
volumes in disklabel. Can you tell me how to set up a raid 1 mirror with
out RAID_AUTOCONFIG please?
Also what do I need in the kernel for RAIDframe with out RAID_AUTOCONFIG?
Is it just pseudo device raid 4 ? Or something like that?
[color=blue]
>
> You don't say what you're running; there was a problem for some
> RAIDframe users on -current between October 10 and November 2 that had
> that very symptom.[/color]
I grabbed the 4.2 sys.tar.gz and I'm running openbsd on a sun blade 100
ultra sparc II system (which is a slow as hell).
[color=blue]
>[color=green]
>> What kind of problems? What error reporting do you get when raidframe
>> detects bad disks?[/color]
>
> Two of my problems have come from running -current.
>
> One of the problems I had was a kernel hang on boot as you've just
> experienced yourself. It occurred after art@'s context switch changes to
> -current, as described above and also here:
>
> [url]http://marc.info/?l=openbsd-misc&m=119289255002826&w=2[/url]
>
> Here's another, caused by disklabel changes for FFS2:
>
> [url]http://marc.info/?l=openbsd-misc&m=118158728932321&w=2[/url]
>
> The only other problem was misunderstanding the limits of
> root-on-raid:
>
> [url]http://marc.info/?l=openbsd-misc&m=115668370718537&w=2[/url][/color]
This is really useful information. Thanks I will take time to read it.
[color=blue]
>
> When RAIDframe fails a disk, it produces a kernel message. I have a cron
> job that checks raidctl -s output, and reports discrepancies. I've had
> two failures caught by RAIDframe, and both times, was alerted by e-mail.
>[/color]
fantastic! Much better than the nasty dell perc IV card I have for
raid in another server then!
[color=blue]
> I've also had a disk problem NOT caught by RAIDframe... caught by
> smartmontools instead.[/color]
Is smartmontools in the ports tree? Is it a crontab thing or a daemon?
[color=blue]
> But that is perfectly understandable, as
> smartmontools caught the error during a long offline test where every
> sector is checked. RAIDframe will only report errors on I/O ops it
> makes.[/color]
Good to hear all your good and bad experiences.
[color=blue]
>
>[color=green]
>>[color=darkred]
>>> --
>>> Replying directly will get you locally blacklisted. Change the address;
>>> use my first name in front of the @ if you want to communicate
>>> privately.
>>>
>>>[/color][/color]
>
> --
> Replying directly will get you locally blacklisted.
> Change the address; use my first name in front of the @ if you want to
> communicate privately.
>
>[/color]
On Wed, 21 Nov 2007 20:09:34 +0000, Khalid Schofield wrote:
[color=blue]
> Yup have RAID_AUTOCONFIG in the kernel. So I was following an example on
> setting up a raid 1 two disk mirror on a web page. It said make wd0d and
> wd1d which will have the OS on it and wd0a and wd1a will have a basic
> system installed so the system will boot. I marked wd0d and wd1d as RAID
> volumes in disklabel. Can you tell me how to set up a raid 1 mirror with
> out RAID_AUTOCONFIG please?[/color]
Autoconfig is used to have the kernel recognize a RAID set at boot time.
Without it, one "configures" the RAID set with raidctl(8) -c after boot
is complete, typically in /etc/rc.conf.local. Autoconfig is most commonly
used for root-on-raid environments, as kernel recognition is required for
it. The raid(4) man page I pointed you to earlier has some detail, and
the raidctl(8) man page has much much more. Both man pages are mandatory
reading.
The OpenBSD Project recommends against using a "howto" document you find
on the web without very careful study and understanding of what each step
is designed to accomplish. Not only are such guides often out-of-date,
they rarely match the reader's environment exactly.
[color=blue]
> Also what do I need in the kernel for RAIDframe with out RAID_AUTOCONFIG?
>
> Is it just pseudo device raid 4 ? Or something like that?[/color]
pseudo-device raid 4. Per the raid(4) man page. If you are using SCSI
devices, you will want to hard code sd(4) devices to specific
target/lun... again, per the man page.
[color=blue]
> I grabbed the 4.2 sys.tar.gz and I'm running openbsd on a sun blade 100
> ultra sparc II system (which is a slow as hell).[/color]
4.2-release, then. Your hanging problem had the same symptoms as mine,
but a different cause. I don't have any specific insights into the cause,
except that autoconfiguration is failing.
[color=blue]
> Is smartmontools in the ports tree? Is it a crontab thing or a daemon?[/color]
Smartmontools is in the tree, and has been for years. It includes a
daemon (smartd) and a control program (smartctl).
I recommend you start slowly... raidctl(8) takes some time to learn how to
use. Rather than starting with a raid-managed system, begin with a single
data partition. Then move on once you've figured out how to
manage the data partition prperly with raidctl -- that you know how to
"fail" a drive, use a hot spare, recover a drive, and the like.
The data partition need not be very large -- when I started via
qemu, I think I used 20MB data partitions, with all sorts of RAID
configurations of RAID 1, 4, 5, and 1/0.
I also use raidctl to manage upgrades -- on a RAID 1 system, I fail one of
the mirrors and confirm the upgrade's success prior to reestablishing the
mirror. One must test out procedures for recovery to the "failed" mirror,
but that was part of my initial testing and education.
--
Replying directly will get you locally blacklisted.
Change the address; use my first name in front of the @ if you want to
communicate privately.