As part of the effort to test integration of LogsDB we would like to run backward compatibility integration tests. Changes in index mode can occur through explicit changes in settings or installation/removal of integrations using different index modes. The index mode operates at the index level and may vary within a single datastream, resulting in coexistence of indices with different index modes. Note that changing the index mode post-index creation isn't feasible. For datastreams' backing indices, the index mode changes only after a rollover operation on a new index.
[ ] Create a new datastream with a standard index as a backing index and index some documents. Change the index mode to “logs” by updating the datastream template and trigger a rollover. Index some other documents. Make sure documents indexed before and after the rollover are correctly indexed and that there is no indexing error. Once indexing is complete the datastream should have two backing indices, one whose mode is “standard” and the other whose mode is “logs”.
[ ] Create a new datastream with a time series index as a backing index and index some documents. Change the index mode to “logs” by updating the datastream template and trigger a rollover. Index some other documents. Notice that, because of the specific rules controlling time series datastreams rollover, a new index won’t be created until the existing time series index end_time has elapsed. As a result, documents indexed after the rollover request is made might still end up in the time series index. Once indexing is complete and enough time has passed the datastream should have two backing indices, one whose mode is “time series” and the other whose mode is “logs”. Make sure documents indexed before and after the rollover are correctly indexed and that there is no indexing error.
[ ] Create a new datastream with a LogsDB index as a backing index and index some documents. Change the index mode to “standard” by updating the datastream template and trigger a rollover. Index some other documents. Make sure documents indexed before and after the rollover are correctly indexed and that there is no indexing error. Once indexing is complete the datastream should have two backing indices, one whose mode is “logs” and the other whose mode is “standard”.
[ ] Create a new datastream with a LogsDB index as a backing index and index some documents. Change the index mode to “time series” by updating the datastream template and trigger a rollover. Index some other documents. Make sure documents indexed before and after the rollover are correctly indexed and that there is no indexing error. Once indexing is complete the datastream should have two backing indices, one whose mode is “logs” and the other whose mode is “time series”.
[ ] Create a new datastream using standard index mode and index some documents. Start a rolling upgrade to a version of Elasticsearch that supports LogsDB and uses default templates so as to use LogsDB as index mode. Make sure documents indexed before and after the rolling upgrade are correctly indexed and that there is no indexing error.
[ ] Create a new datastream using standard index mode and index some documents. Start a full cluster upgrade to a version of Elasticsearch that supports LogsDB and uses default templates so as to use LogsDB as index mode. Make sure documents indexed before and after the full cluster upgrade are correctly indexed and that there is no indexing error.
Description
As part of the effort to test integration of LogsDB we would like to run backward compatibility integration tests. Changes in index mode can occur through explicit changes in settings or installation/removal of integrations using different index modes. The index mode operates at the index level and may vary within a single datastream, resulting in coexistence of indices with different index modes. Note that changing the index mode post-index creation isn't feasible. For datastreams' backing indices, the index mode changes only after a rollover operation on a new index.