Re: Testers wanted: Gvinum patches of SoC 2007 work

On man, sep 17, 2007 at 04:18:42 -0700, Barnabas wrote:
> 
> Hi Ulf
> 
[...]
> I have had a lot of issues with the semi-finished version of gvinum that is
> currently available though the freebsd distro. I have had a lot of DMA
> read/write timeout issues lately. Especially under heavy load. But only on
> my ata drives - the scsi drives runs perfectly. First I thought it was a
> dead disk, but I have been seeing the issue on my new raid5 setup as well.
> The same issue on two disk at the same time is not very likely. If you have
> any info on how I am to patch the current vinum driver with the new changes
> you made it would be great. 
> 
> I really appreciate someone is looking at this exellent software. It has not
> really worked 100% since the port to geom was started.
> 
> Heres my config:
> 
> FreeBSD sauron.barnabas.dk 6.2-RELEASE-p6 FreeBSD 6.2-RELEASE-p6 #0: Wed Jul
> 18 23:33:58 CEST 2007    
> root_at_sauron.barnabas.dk:/usr/src/sys/i386/compile/KERNEL_6_2  i386
> 
> Here is some of the errors I have experienced:
> 
[...]
> Sep 17 16:54:03 sauron kernel: subdisk10: detached
> Sep 17 16:54:03 sauron kernel: ad10: detached
> Sep 17 16:54:03 sauron kernel: ad11: FAILURE - device detached
> Sep 17 16:54:03 sauron kernel: subdisk11: detached
> Sep 17 16:54:03 sauron kernel: ad11: detached
Hi,

I think this problem is not related to gvinum actually. As you can see, there
are detach notices from ad11. This is most probably a problem with the hd
controller, or the controller driver, or the drives (but highly unlikely since
both fail), or a bug in the driver (since this only happens under heavy load).

You can try to use atacontrol(8) to set the transfer mode of the ATA controller.
This might indicate more specifically what's the problem. Other than that,
giving information about your controller and harddrive hardware would help.

Also, because of the ATA problem , applying the new gvinum patch wouldn't help
you solve any of these issues. 

However, since you tried the patch and it failed (as you noted in the e-mail you
sent privately), I'd like to know the output you got from that (dmesg, 'gvinum
list', 'gvinum printconfig').

-- 
Ulf Lilleengen