WIP - Support for facilitating creating a read-side Cassandra table [don't merge] #130
Conversation
As highlighted in lagom/lagom-java-chirper-example#24, making sure that a read-side table is (eventually) created can be challenging. In particular, users need to ensure that the create table statement is retried if Cassandra isn't available. This commit addresses the problem by encapsulating the retry logic in a new API class: `CassandraReadSideTableCreator`. Both the number of attempts and the interval between each attempt is configurable.
|
Here is an example demonstrating how this would work https://github.com/dotta/activator-lagom-scala-chirper/blob/issue/cassandra-connection/chirp-impl/src/main/scala/sample/chirper/chirp/impl/ChipServiceImpl.scala#L45-L50 |
| }.map(_ => Done) | ||
|
|
||
| createTable.onFailure { | ||
| case err: Throwable => log.error(s"Failed to execute create table statement:\n$createTableStatement. Reason: ${err.getMessage}", err) |
There was a problem hiding this comment.
newlines should be avoided in log messages, in my opinion
|
looks like a good utility |
|
I don't think table creator is a good name - there are more things than creating tables that you might want to do on startup, you might want to update tables for example. I also think it would be better to provide a utility that would be able to ensure that the table creation is successful before other events are processed. While waiting for the table creation statement to be successful, other events submitted might go into a bounded queue, and would time out if they sat in the queue for too long. |
|
Agreed. Then this utility need to be reworked into something more general. Closing for the moment. |
I've opened this PR to collect feedback on the functionality and its current implementation
As highlighted in lagom/lagom-java-chirper-example#24, making sure that a
read-side table is (eventually) created can be challenging. In particular, users
need to ensure that the create table statement is retried if Cassandra isn't
available.
This commit addresses the problem by encapsulating the retry logic in a new API
class:
CassandraReadSideTableCreator. Both the number of attempts and theinterval between each attempt is configurable.
@patriknw what do you think?