Migrating to new hardware, questions about Data Node / Opensearch

Graylog Central (peer support)
1

I’m currently running a single server with graylog 6.2, mongodb 7 and opensearch 2.15 all on a the same physical box. It’s working fine for me, but the hardware is aging and I’d like to replace it. I’ve got the new machine set up with the same versions of everything installed but had some questions about possible ways to migrate to the new box, as well possibly migrating to Data Node during or after the migration.

I’m currently planning on snapshotting the existing opensearch instance to shared storage and then restoring on to the new server following this guide, then moving mongodb and all config files, and then just sending it.

  • I know running graylog and data node isn’t recommended (and neither is running es/opensearch on it), but I’ve been running one piece of hardware for a few years and it’s working fine and I’d like to avoid buying a second piece of hardware. Is it possible to safely install to DataNode on the same hardware as graylog/mongodb for a small setup?
  • If it is possible, should I restore my opensearch snapshot to a self managed opensearch on the new server, then migrate that to DataNode, or should I migrate the old server to DataNode, then migrate that to the new server?
  • Is there a better way to do this? (Like, adding both servers to a cluster, then disable the old one and let data age out?)

Thanks!

2

Hi @rockkoh
Technically there is nothing that would prevent you to run all services on one machine. It’s not recommended production setup, as there is no failover, no redundancy and the services will compete for resources. But if you have a really small setup and everything is working for you, then you can continue with such setup, you won’t have any problems with datanode either.

Mixed cluster of plain opensearch and datanode/opensearch is not supported and I wouldn’t go that route.

should I restore my opensearch snapshot to a self managed opensearch on the new server, then migrate that to DataNode

I think this will be the easiest approach, as you have a bit more flexibility with plain opensearch. It should be rather straightforward process. But if you hit any problems, feel free to report them here.

Best regards,
Tomas