Interactive update command improvements

zalando-stups / planb-cassandra

Plan B Cassandra for STUPS/AWS with static IPs

Other

27 stars 18 forks source link

Interactive update command improvements #203

Open a1exsh opened 6 years ago

a1exsh commented 6 years ago

We would like to make the update command behavior less surprising by implementing the following improvements:

[ ] document how to resolve 'failed' state, maybe provide a hint as well
[ ] assess the current state of the node and allow to skip updating if it already matches the target configuration
[x] unattended mode: fire and forget (should still stop if some nodes go down unexpectedly)
[x] better node liveliness probe (DownEndpointCount is not a good fit)

lmineiro commented 6 years ago

If the original node has detailed monitoring enabled it will be gone in the new node

a1exsh commented 6 years ago

@lmineiro didn't know about that. We should probably have it as an option for create and update.

a1exsh commented 4 years ago

skip updating if it already matches the target configuration

In common cases like updating the AMI, this can be addressed by using filters. That is filter for nodes running the old AMI. This makes it easier to continue if update broke in the middle.