From mboxrd@z Thu Jan  1 00:00:00 1970
From: artur.paszkiewicz@intel.com (Artur Paszkiewicz)
Date: Wed, 1 Jun 2016 10:23:37 +0200
Subject: RAID0 mdadm Question
In-Reply-To: <CA+Tq-Rr9yU1VNr-XQ=S5_hnP0saZz7poc8SCESaidZ43Wmak+w@mail.gmail.com>
References: <CA+Tq-RoTOSwAbJJFBgkDEC8LV-CiFOHKvbC3i3tkUUoxu03ziQ@mail.gmail.com>
 <574D9790.1080106@intel.com>
 <CA+Tq-Rr9yU1VNr-XQ=S5_hnP0saZz7poc8SCESaidZ43Wmak+w@mail.gmail.com>
Message-ID: <574E9B89.60109@intel.com>

On 06/01/2016 04:08 AM, Hiroyuki Sato wrote:
> Hello Artur
> 
> Thank you for your replying.
> 
> I found /dev/md126. But It does not contain partition information.
> I'm not sure why partition information lost.
> 
> Could you tell me if you know any other commands?
> 
> Best regards.
> 
> 
> Step1 scan drives
> 
>   mdadm --assemble --scan
>   mdadm: Container /dev/md/imsm0 has been assembled with 2 drives
>   mdadm: Started /dev/md/0_0 with 2 devices
> 
> 
>   # cat /proc/mdstat
>   Personalities : [raid0]
>   md127 : inactive nvme0n1[1](S) nvme1n1[0](S)
>         6306 blocks super external:imsm
> 
>   md126 : active raid0 nvme1n1[1] nvme0n1[0]
>         781416448 blocks super external:/md127/0 128k chunks
> 
>   unused devices: <none>
> 
> Step2 check partition information
> 
>   parted /dev/md126
>   GNU Parted 3.1
>   Using /dev/md126
>   Welcome to GNU Parted! Type 'help' to view a list of commands.
>   (parted) p
>   Error: /dev/md126: unrecognised disk label
>   Model: Linux Software RAID Array (md)
>   Disk /dev/md126: 800GB
>   Sector size (logical/physical): 512B/512B
>   Partition Table: unknown
>   Disk Flags:
> 
> 2016-05-31 22:54 GMT+09:00 Artur Paszkiewicz <artur.paszkiewicz at intel.com>:
>> On 05/31/2016 11:41 AM, Hiroyuki Sato wrote:
>>> Hello
>>>
>>> NVMe Newbie question.
>>>
>>> I have two P3600 PCIe NVMe Cards.
>>> And I created RAID0 volume with mdadm command.
>>> It works fine. But after reboot, I can't mount file system it.
>>> It seems lost partition table.
>>> (No partition information)
>>>
>>> Environment
>>>   - NVMe: Intel P3600 cards * 2
>>>   - Linux: 4.4.0
>>>   - OS: CentOS7
>>>
>>> Question
>>>
>>>   1, Do I need re-create file system on each Boot time?
>>>
>>>   2, If not What step is missing?
>>>
>>>     generate mdadm.conf?
>>>
>>> Best regards.
>>>
>>> NVMe RAID step
>>>
>>>   Basically I followed this doc
>>>     https://communities.intel.com/community/itpeernetwork/blog/2015/10/01/how-to-use-and-benchmark-nvme-ssd-create-a-software-raid-and-analyze-performance-the-answers-are-here
>>>
>>>   Step1: create container
>>>
>>>     mdadm -C /dev/md/imsm0 /dev/nvme0n1 /dev/nvme1n1 -n 2 -e imsm -f
>>>     mdadm: /dev/nvme0n1 appears to be part of a raid array:
>>>            level=raid0 devices=0 ctime=Thu Jan  1 09:00:00 1970
>>>     Continue creating array? y
>>>     mdadm: container /dev/md/imsm0 prepared.
>>>
>>>   Step2, create md device
>>>     mdadm -C /dev/md0 /dev/md/imsm0 -n 2 -l 0 -c 128 -f
>>>     mdadm: array /dev/md0 started.
>>>
>>>   Step3: Create Partition
>>>     parted /dev/md0
>>>
>>>     (parted) mkpart
>>>     Partition name?  []?
>>>     File system type?  [ext2]? xfs
>>>     Start? 0%
>>>     End? -1
>>>
>>>     (parted) p
>>>     Model: Linux Software RAID Array (md)
>>>     Disk /dev/md0: 800GB
>>>     Sector size (logical/physical): 512B/512B
>>>     Partition Table: gpt
>>>     Disk Flags:
>>>
>>>     Number  Start   End    Size   File system  Name  Flags
>>>      1      1049kB  800GB  800GB
>>>
>>>     (parted) quit
>>>     Information: You may need to update /etc/fstab.
>>>
>>>   Step4: Newfs
>>>
>>>     /sbin/mkfs.xfs -K /dev/md0p1 -f
>>>
>>>   Step5: mount
>>>
>>>     mount -o noatime,nodiratime,nobarrier /dev/md0p1 /mnt/nvme1

I don't see a step where you create a partition table, so maybe it's
reusing a gpt table that was present on the first device and that is
causing problems. Please try writing a fresh partition table between
steps 2 and 3 like this:
# parted /dev/md0 mklabel gpt