2021-08-16 19:17:24: zhaowei@zettadb.com created the issue
Currently the add_shards.py, add_computing_node.py, create_cluster.py scripts are not idempotent --- if they are used improperly and installation fails in the middle, it's hard to resume the installation or even simply restarting the installation --- we need to achieve idempotency of such scripts so that users can very easily resume the installation.
this is needed even if we have one-stroke cluster installation tools because users may want to add more storage shards or computing nodes after an cluster has been running for a while.
To achieve idempotency, either undo intermediate effects of such install scripts when installation fails, or find the right phase/point to resume the installation. But do not delete any storage node(s), meta-shard node(s) or existing computing node(s).
Issue migrated from trac ticket # 137 www.kunlunbase.com
component: cluster provision | priority: major
2021-08-16 19:17:24: zhaowei@zettadb.com created the issue