RE: invalid page header on postgres using gpt/dkctl

[Derrick Lobo <derrick.lobo%givex.com@localhost>]

I destroyed the dkctl wedges and destroyed sd0 .. and the error did not go
away .. it seems the bios is diff from what netbsd sees.. 


sd0 at scsibus0 target 0 lun 0: <LSI, MR9240-8i, 2.70> disk fixed
sd0: fabricating a geometry
sd0: 5585 GB, 5719182 cyl, 64 head, 32 sec, 512 bytes/sect x 11712884736
sectors
sd0: fabricating a geometry
sd0: mbr partition exceeds disk size
sd0: tagged queueing

Derrick Lobo
-----Original Message-----
From: J. Hannken-Illjes [mailto:hannken%eis.cs.tu-bs.de@localhost] 
Sent: Tuesday, June 03, 2014 10:44 AM
To: Derrick Lobo
Cc: port-amd64%NetBSD.org@localhost; netbsd-users%NetBSD.org@localhost;
netbsd-ports%netbsd.org@localhost
Subject: Re: invalid page header on postgres using gpt/dkctl

On 03 Jun 2014, at 16:32, Derrick Lobo <derrick.lobo%givex.com@localhost> wrote:

> Hi All
> 
> I get a "invalid page header" on postgres .
> 
> The servers uses 3 x 2Tb drivers and I used gtp/dkctl to configure a 
> disk.. The server has a LSI controller and the DISK are SATA.
> 
> I DO NOT see any error messages on dmesg or /var/log/messages.... 
> however my database keeps getting corrupt..

Sure? See below ...

> Using dkctl I am unable to fix or find any badsectors.
> 
> The commands used to configure the disk was
> 
> # gpt show sd0
>        start         size  index  contents
>            0            1         PMBR
>            1  11714691071
> 
> ### Create a new wedge
> # gpt create sd0
> # gpt show sd0
>        start         size  index  contents
>            0            1         PMBR
>            1            1         Pri GPT header
>            2           32         Pri GPT table
>           34  11712884669
>  11712884703           32         Sec GPT table
>  11712884735           1         Sec GPT header
> # gpt add -t ufs -s 11712884669 sd0
> # dkctl sd0 addwedge dk0 34 11714691005 ufs

Here 34+11714691005 -> 11714691039 is greater than the disk size of
11712884736

> dkctl sd0 listwedges
> # dkctl dk0 getwedgeinfo
> dk0 at sd0: dk0
> dk0: 11714691005 blocks at 34, type: ufs # gpt show sd0
>        start         size  index  contents
>            0            1         PMBR
>            1            1         Pri GPT header
>            2           32         Pri GPT table
> 
>           34  11714691005      1  GPT part - NetBSD UFS/UFS2
>  11714691039           32         Sec GPT table
>  11714691071            1         Sec GPT header
> 
> reboot
> 
> ## Create new filesystem and mount it
> # newfs -O2 /dev/rdk0
> # mount /dev/dk0 /mnt
> 
> Dmesg for Sd0
> 
> sd0 at scsibus0 target 0 lun 0: <LSI, MR9240-8i, 2.70> disk fixed
> sd0: fabricating a geometry
> sd0: 5585 GB, 5719182 cyl, 64 head, 32 sec, 512 bytes/sect x 
> 11712884736 sectors
> sd0: fabricating a geometry
> sd0: mbr partition exceeds disk size

I would call this an error message.

> sd0: GPT GUID: c4b3e41f-d14b-11e3-950f-00259084a564
> dk0 at sd0: c4b3e43c-d14b-11e3-950f-00259084a564
> sd0: tagged queueing
> 
> Derrick Lobo

--
J. Hannken-Illjes - hannken%eis.cs.tu-bs.de@localhost - TU Braunschweig 
(Germany)