0

I have a large (2.2PB, ~6 Billion files) EC pool used by radosgw. It is still under very heavy use by users for reading and writing. However, I want to start using a new ceph pool and migrate all the data in this ceph pool to the new one.

Unfortunately, taking a downtime is not going to be an option as time sensitive data is actively stored and retrieved from this pool and the downtime would likely have to be days to migrate this much data.

Is there a way to migrate this pool without a significant downtime? The closest discussion I found is in this thread where they talk about setting up a new default pool but it doesn't have enough details for my skill level as to how to implement the solution.

Any instructions on how I could accomplish this migration would be appreciated.

OS Release: Ubuntu 20.04
Ceph Release: Octopus

user2074364
  • 31
  • 3
  • Ceph is not such an easy solution where you get a step-by-step instruction for such a complex task. You should familiarize yourself with ceph to the necessary skill level so you understand the thread you mentioned. And if you are at that skill level start testing heavily before touching the production environment. – eblock Jun 04 '21 at 05:38
  • What's the need of migrating to a new pool? – Seena Fallah Jun 30 '21 at 16:51
  • @SeenaFallah To change the erasure profile for higher redundancy – user2074364 Jul 01 '21 at 17:31
  • @user2074364 why not simply create a new crush rule and change the pool crush rule? – Seena Fallah Jul 01 '21 at 20:47
  • Thanks! I wasn't aware you could do that. I thought a pool's erasure profile couldn't be changed once you selected a k/m combination. Out of curiosity I'd still love to know the answer to the original question even if I won't need it. – user2074364 Jul 02 '21 at 15:24

0 Answers0