Open bheni opened 1 year ago
I think is probably caused by the second sql server coming up as a primary and the cluster entering detected_broken_config
state. The actual cluster role is stored in a persisted SQL variable, and my understanding is that those are typically shared between sql-server
invocations in different directories because they're stored in HOME/DOLT_ROOT_PATH.
This might also be a particular quirk of how cluster replication stores the persistent role and epoch. I'm not sure if it stores it the exact same way as the default SQL path for persistent global variables does.
I was able to get this working by setting DOLT_ROOT_PATH to a different directory for each running server process.
Setup the directories and configs:
Then in two separate terminal windows with one in the primary directory and one in the standby1 directory run
dolt sql-server --config config.yaml
In a third terminal window connect to the primary and try to create a database and table:If you log into standby1 instead of primary you will now see that the empty database "db" was replicated.
I don't know if I have a config issue, but even if I do the behavior is not great.