Backing up etcd without stopping

[David Caldwell]

Hey guys,

I have a couple of -- hopefully simple! -- questions (assuming a cluster of three etcd nodes):

1. Is there a safe and supported way of backing up etcd without taking it offline?

2. If I *do* perform a backup without stopping etcd, is there a possibility that the backup will be unusable if there is a write to the etcd database whilst taking the backup?

3. If I stop 1 etcd server only and perform the backup on that host, is this the best way even though I potentially miss data written to the other nodes whilst the backup is in progress?

I have read what documentation I can find but I cannot find a definitive answer to the best backup method.

Thanks,

David.

[anthony...@coreos.com]

1. Yes; v3 snapshots. https://github.com/coreos/etcd/blob/master/etcdctl/README.md#snapshot-subcommand
2. v3 snapshots are transactions; there shouldn't be corruption from interleaved writes.
3. It shouldn't be necessary to offline the cluster.

You may also be interested in the recovery/backup guide:
https://github.com/coreos/etcd/blob/master/Documentation/op-guide/recovery.md