When Cassandra is temporarily down make sure the connection is resumed about kong HOT 4 CLOSED

#28 now uses the cosocket API. So we might loose some queries but the connection will be resumed as soon as cassandra is reachable again. Now we need to know if this doesn't throw any unexpected errors that could disturb the execution of plugins not using the DB.

from kong.

Right now this doesn't happen. Connection to Cassandra is never resumed and nginx logs the following error:

2015/02/24 00:35:42 [error] 19605#0: *1 [lua] access.lua:28: execute(): Cassandra returned error (Unprepared): "Prepared query with ID 29083996e2fe585226bc6e89ae37c690 not found (either the query was not prepared on this host (maybe the host has been restarted?) or you have prepared too many queries and it has been evicted from the internal cache)", client: 127.0.0.1, server: , request: "GET / HTTP/1.1", host: "localhost:8000"

from kong.

if the cluster shuts down and restarts, prepared queries are dropped and need to be re-prepared.

from kong.

BTW, re-reading the issue: it won't cause any issue if cassandra is unreachable because of network issues beyond dropping the queries.

It will only be a more critical issue if cassandra was restarted for some reason, because the statements aren't prepared anymore, resulting in errors and no query will succeed.

from kong.