[DPE-7641] Authentication #8
Conversation
Did I understood it correctly, that this question is about secret changed event? I thought I'll put it in another PR, but considering that this PR is relatively small, I'll include it here. Thanks! |
Zvirovyi
requested review from
welpaolo,
deusebio and
marcoppenheimer
and removed request for
welpaolo,
deusebio and
marcoppenheimer
| seeds=["127.0.0.1:7000"], | ||
| authentication=False, | ||
| ) | ||
| self.workload.start() |
There was a problem hiding this comment.
question If I understand the logic correctly, here we are starting the server with no authentication. This seems weird and it could pose a security threat. I understand that we are possibly restricting the listen_address into localhost, but why do we need to do this?
Another possibility could also be to start the server using a temporary username and password that only the units knows, such that it can still use the service in authenticated fashion, and then change the password to use the "official" one.
anyhow, let's discuss later at daily
| keystore_password=self.state.unit.keystore_password, | ||
| truststore_password=self.state.unit.truststore_password, | ||
| authentication=bool(self.state.cluster.cassandra_password_secret), |
There was a problem hiding this comment.
quesiton why do we allow this? Maybe I would not expose this, and only allow authentication to be true, just to prevent mistakes. We must ensure that authentication is enabled at all times, if external access is possible.
| yaml.dump(config, allow_unicode=True, default_flow_style=False) | ||
| yaml.dump( | ||
| self._merge_dicts( | ||
| [ |
There was a problem hiding this comment.
praise I really really really like how you splitted the config into separate section. It is a lot more maintanable IMHO
| self.database_manager.update_system_user_password( | ||
| "cassandra", self.state.cluster.cassandra_password_secret | ||
| ) | ||
| self.cluster_manager.prepare_shutdown() |
There was a problem hiding this comment.
question why do we drain before reconfiguring the password? The drai operation may be massive?
| ): | ||
| self.database_manager.update_system_user_password("cassandra", password) | ||
| self.state.cluster.cassandra_password_secret = password | ||
| self.cluster_manager.prepare_shutdown() |
There was a problem hiding this comment.
issue it is strange that we need to drain for every config change/rotation password
Zvirovyi
force-pushed
the
authentication
branch
from
f923544 to
7476eea
Compare
Overview
This PR:
cassandrauser. Showcase diagram of changes is presented below.Diagram of startup logic changes
Diagram describes
on_startevent handling represented from the leader perspective as there's no changes for the subordinate units.