Re: best way to replace all disks

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Il giorno mar 23 apr 2019 alle ore 03:09 Peter Grandi
<pg@xxxxxxxxxxxxxxxxxxxxxx> ha scritto:
> 2TB is nowadays a pretty minuscule server (a number of people I
> know have in their laptop one or two 2TB NVME drive...), but
> then 100Mb/s is small in relation to that.

Sure, but as wrote, this is a very old server, with very old hardware
on a very old network
that I can't change easily.

Yes, a huge upgrade is planned, but that's impraticable in next few days.

> Except that your situation is that you have 2x RAID1 sets each
> with 2x 1TB disks, the first 2 disks have 4 partitions and each
> partition is mirrored on the other disk.
>
> Since the two pairs of disks are independent, the obvious way to
> proceed is to remove completely one of the two disk pairs, and
> use the two slots thus freed to upgrade the RAID sets on the
> first disk pair; then do the other.
>
> So for example, assuming that disk A1 is 'sda', disk 'B1' is
> 'sdb', disk 'C1' is 'sdc', disk 'D1' is 'sdd', and you have
> new disks A2, B2, C2, D2:
>
>   * Remove C1 and D1.
>   * Put in their place A2 and B2.
>   * Expand A1 and B1 into A2 and B2.
>   * Remove A1 and B1, A2 and B2.
>
>   * Put in their place C1 and D1, C2 and D2.
>   * Expand C1 and D1 into C2 and D2.
>   * Remove C1 and D1.
>   * Put in their place A2 and B2.
>
> To expand A1 and B1 into A2 and B2 (and similarly for the other
> pair), the simplest and safest way is:
>
>   * Make the RAID set on A1 and B1 read-only.
>
>   * Configure A2 and B2 with the partitions you want.
>
>   * Configure those partition pairs each as a RAID1, and format
>     each RAID1 set with the filesystems of choice, and mount them.
>
>   * Copy using 'rsync -axAXHO' the data from the existing RAID
>     sets to the new RAID sets, e.g. from '/mnt/md-old[0-3]/' to
>     '/mnt/md-new[0-3]/'.

The total size of all arrays is less than 2TB. So, you are probably
right (if I understood properly)
that I can use a single 2TB disk to hold all arrays (that would be a
much better solution)

So:
- power off the server
- remove 1 disks from each mirror
- put one huge brand new disk
- add some partitions to this disk
- use these new partitions as replacement for the older ones stored on
removed disks.
- let the raid rebuild itself [1]
- when rebuild is finished, start from scratch by replacing the other old disk

[1] As wrote multiple times, i have to be 100% that nobody is changing
the underlaying array during the whole operation, or the removed disk
would be out-of-sync and useless in case of disaster
Booting from a live image (which image do you suggest), without
mounting the filesystem stored on the RAID, would be enough, right ?
Is a brand-new mdadm able to handle properly a raid created on 2007?
The newly synced array, would be usable by the 2007 mdadm when server
boot up again?



[Index of Archives]     [Linux RAID Wiki]     [ATA RAID]     [Linux SCSI Target Infrastructure]     [Linux Block]     [Linux IDE]     [Linux SCSI]     [Linux Hams]     [Device Mapper]     [Device Mapper Cryptographics]     [Kernel]     [Linux Admin]     [Linux Net]     [GFS]     [RPM]     [git]     [Yosemite Forum]


  Powered by Linux