Does experimental feature `consistent-topology-changes` support online modification?

ScyllaDB
1

Now I have a up and normal 5.4.0 cluster.

image
image1176×279 10.8 KB

After adding configuration consistent-topology-changes to them, only the seed node is functioning properly. The other two nodes are not displayed in the cluster, but their logs do not report any errors.
image
image1459×304 15.8 KB

However, the logs of the seed node show that the other two nodes are up and normal.

INFO  2024-01-16 06:17:27,366 [shard 0:main] raft_group0 - gaining leadership
INFO  2024-01-16 06:17:28,302 [shard 0:goss] gossip - InetAddress 100.71.8.19 is now UP, status = NORMAL
INFO  2024-01-16 06:17:28,328 [shard 0:goss] gossip - InetAddress 100.71.8.18 is now UP, status = NORMAL
INFO  2024-01-16 06:17:28,495 [shard 0:comp] compaction - [Compact system_distributed.cdc_generation_timestamps ebc9ab00-b436-11ee-9ea8-adab234ee8bf] Compacted 2 sstables to [/var/lib/scylla/data/system_distributed/cdc_generation_timestamps-fdf455c4cfec3e009719d7a45436c89d/me-3gct_0hh3_1f0f42euxqvtiwkmv3-big-Data.db:level=0]. 11kB to 5kB (~50% of original) in 1137ms = 9kB/s. ~256 total partitions merged to 1.
2

Hi,

The consistent-topology-changes experimental feature does not yet support upgrades. I.e. if you start with a cluster without the feature, and then try to enable it, the cluster will end up in some undefined state.

To get a functioning cluster with consistent-topology-changes, you need to bootstrap it from scratch with this option.

But remember that this is an experimental feature, it’s not production ready. It’s under heavy development with backward incompatible changes happening every week. It’s fine to play with it with clusters that you’re going to discard.

1 Like