Pool migration after Perc H700 to H200 upgrade

[Rapper_skull]

Hi,
I have a Dell PowerEdge T610 system, currently running FreeNAS 11.2, configured with a single pool spanning 6 disks and a spare. Unfortunately the system has a Perc H700 RAID controller and I configured 7 RAID 0 virtual disks of 1 physical disk each to make everything work. Since I would like to replace the controller with a Perc H200 (properly crossflashed) I was wondering how to migrate the pool.
Is it possible to just replace the controller, leave the disks unconfigured, then re-import the pool in FreeNAS? Or do I need to backup all the data and create the pool from scratch?
Thank you in advance for your help.

 

[melloa]

create the pool from scratch

The above. ZFS won't read the pool from the raid controller.

 

[Rapper_skull]

The above. ZFS won't read the pool from the raid controller.

But can you do it without losing data?

 

[jgreco]

Not really possible, assuming that it's done the usual "create its own little proprietary partition system on the disk" thing that so many RAID controllers do.

https://www.ixsystems.com/community...s-and-why-cant-i-use-a-raid-controller.81931/

I discuss this briefly at the end of point #5. What's happened is that the controller makes a partition that encapsulates your JBOD virtual device, so the data that OUGHT to be at LBA #0 has been moved forward into an LBA a little further in to the disk, and the JBOD's proprietary partition is a wee bit smaller than the physical disk. It would theoretically be possible to simply relocate the data (reading it maybe a gigabyte at a time and rewriting it where it ought to be) which would transform the disk into something usable. I'm not aware of any tools that would allow an end-user to do this without getting down and dirty with stuff, though.

 

[Rapper_skull]

Not really possible, assuming that it's done the usual "create its own little proprietary partition system on the disk" thing that so many RAID controllers do.

https://www.ixsystems.com/community...-and-why-can't-i-use-a-raid-controller.81931/

I discuss this briefly at the end of point #5. What's happened is that the controller makes a partition that encapsulates your JBOD virtual device, so the data that OUGHT to be at LBA #0 has been moved forward into an LBA a little further in to the disk, and the JBOD's proprietary partition is a wee bit smaller than the physical disk. It would theoretically be possible to simply relocate the data (reading it maybe a gigabyte at a time and rewriting it where it ought to be) which would transform the disk into something usable. I'm not aware of any tools that would allow an end-user to do this without getting down and dirty with stuff, though.

Thank you for your very datailed answer. One thing that makes me thinking is that I can actually see the type of device running camcontrol devlist. I guess the only way to know is to connect one hard drive to a PC and analyze the raw content. But at this point, what should I look for? Can I identify LBA #0 by searching for the MBR?
Thank you for your help

 

[jgreco]

Yes, you really have to look at what's happened on the disk. Some RAID controllers do the partition thing, especially common for "virtual disk" JBOD's, but sometimes you can luck out and they left the disk readable and instead stole disk space at the back of the disk.

It won't HURT to hook the disks up to an HBA and then ask FreeNAS to import the pool, by the way. Of course if you do anything beyond that, your data's at risk, and it's worth having a backup.

 

[Rapper_skull]

Yes, you really have to look at what's happened on the disk. Some RAID controllers do the partition thing, especially common for "virtual disk" JBOD's, but sometimes you can luck out and they left the disk readable and instead stole disk space at the back of the disk.

It won't HURT to hook the disks up to an HBA and then ask FreeNAS to import the pool, by the way. Of course if you do anything beyond that, your data's at risk, and it's worth having a backup.

Perfect. But let's suppose I have to create the pool from scratch: is there a simple way to backup and restore everything, without having to create all the datasets and zvols again (and above all without affecting jails and VMs)?

 

[Patrick M. Hausen]

If you can install both controllers into the same chassis simultaneously, you could probably move one disk at a time with resilvering after every disk moved.

 

[hervon]

If you can install both controllers into the same chassis simultaneously, you could probably move one disk at a time with resilvering after every disk moved.

I don't think this makes sense. He has to create a new pool. Best way I see is with the ZFS send command from old to new pool. Use the -R flag to preserves all properties.

 

[Rapper_skull]

If you can install both controllers into the same chassis simultaneously, you could probably move one disk at a time with resilvering after every disk moved.

I also thought about it. Since my pool is composed of 3 two-disk mirrors I could remove half of the disks and keep them in the top row of the backplane, while connecting the new controller to the second half. Now it should be possible to resilver the pool. After doing that I could connect the top half of the backplane to the new controller too and repeat.

 

[Dan Tudora]

hello
make a backup of data first is not so bad
ceers

 

[Dan Tudora]

hello
can do that with a HDD in H200 controller (with suficient space) and ZFS send and receive command for the entire pool to have another copy of data
just in case :D
succes

 

[Rapper_skull]

Thank you all. Just orderer my PERC H200, I'll let you know how it goes.

 

[Rapper_skull]

I'm also sharing this post I found on Reddit, in case it could help someone else. It seems that the RAID information is stored at the end of the disk, and it should be possible to just import the pool.

 

[Patrick M. Hausen]

Anyway afterwards have a look at the GPT partition tables and probably use gpart recover.

 

[Rapper_skull]

OK, today I removed the PERC H700 and installed the H200, which I previously crossflashed to P20 IT firmware. I only got some warnings saying the secondary GPT header is not in the last LBA, but gpart successfully resolved the issue.
Unfortunately I have another problem: the A and B ports are swapped: drives 0 to 3 became 4 to 7 and vice-versa (actually, since I have an empy slot, the order is a little bit messed up). Swapping the cables gives me an error about missing A port cable. As I understand it is a common issue, but I'm not sure it affects every H200 unit. Does anyone have any idea?

 

[hervon]

Normally it should not affect your pool (or anything as a matter of fact). Does it ?

 

[Rapper_skull]

Normally it should not affect your pool (or anything as a matter of fact). Does it ?

No, it doesn't, but the disks are reported in a very strange order and it makes very difficult to identify them.

 

[Redcoat]

the disks are reported in a very strange order and it makes very difficult to identify them

There's a script here that you might like https://www.ixsystems.com/community...ktype-serial-num-multipath.59319/#post-421424

 

[Rapper_skull]

There's a script here that you might like https://www.ixsystems.com/community...ktype-serial-num-multipath.59319/#post-421424

Thank you, but for now I just decided to phisically switch the HDDs, remembering that the top and bottom row are switched, and it should be easy to track down the disks.