An incremental strategy for large amounts of storage

From FnordWiki
Revision as of 04:49, 11 July 2012 by Adj (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

Along with the PE1950s, I've picked up a PowerEdge 2950. This is a 2U machine with 6 drive bays. The current situation with a pair of 500 Gibyte drives mirrored leaves a bit to be desired in terms of available room. So here's a plan to get to ~12Tibytes of usable space:

  • Acquire a pair of 3Tibyte SATA drives, ideally from different manufacturers.
  • Do an OS install as before, with a mirrored 1Gibyte /boot and a mirrored 1Gibyte / filesystem.
  • Create a third partition on each drive covering approximately 80% of remaining space. (The drives are being purchased piecemeal and will not all have exactly the same sector counts.)
  • Get backups running -- there will be a period here where a one disk failure will erase everything.
  • Using third partition on these drives, create a three disk MD RAID5 with a missing member. This device should have about 4.8Tibytes of usable space.
  • Acquire third drive, extend mirrors of /boot and / onto it. (2 disk RAID1 becomes a 3 disk RAID1.) Create third partition of size greater or equal to size of existing RAID5 members. Add to RAID 5 array.
  • Enjoy protection against single disk failures. But continue doing backups.
  • Acquire fourth drive. Partition as above. Re-shape RAID5 into RAID6. Reshape 3 member RAID1 arrays into 4 member RAID1 arrays.
  • Enjoy protection against two disk failures. Continue doing backups.
  • Acquire fifth drive. Partition as above. Re-shape 4 member RAID6 array into 5 member RAID6 array.
  • Enjoy increased space.
  • Acquire sixth drive. Partition as above. Re-shape 5 member RAID6 array into 6 member RAID6 array. Reshape 5 member RAID1 arrays into 6 member RAID1 arrays.
  • Enjoy increased space. Continue doing backups.
  • Fail (mdadm /dev/mdX --fail /dev/sdYn) all array members on a single physical disk.
  • Repartition disk, growing MD member partition to end of disk.
  • Add re-partitioned disks's MD member partitions back into RAIDs they came from. Allow array resynchronization to complete.
  • Repeat process for additional physical disks.
  • mdadm --grow --size=max to grow the finished array to its maximum size (as determined by size of member partitions. Should be really close to 12Tibytes.)
  • Continue doing backups. Collect large amounts of pr0n. Get more backup capacity.

End result: 12Tbytes of space (24x the current situation.) Can withstand two disk failures. Probably faster. Lots of room to play with different, shiny filesystems. And for getting all of the camcorder video onto a computer, instead of on HDV tapes.