I have read a number of posts on the issue still not have answers to my questions.
I want to use ES in a scenario where:
- new and updateable data is replicated to ES indexes from the SQL database (when changes occur)
- old, read only data is stored in ES indexes only
I understand that ES may loose data (in some rare cases). I would like to set up some safe framework to handle those rare cases to be sure the data is not lost.
Questions I have are:
- are there any other use cases appart from the unavailability/crashing of cluster nodes, when the data may be lost? Can data be lost if all cluster nodes are available and working normally?
- is there a method I can safely detect each node unavailability/crash? Is every crash/unavailability logged?
- can data be lost for a read-only index? E.g. - while moved from one node to another because of loosing updates to cluster status?
Will be gratefull for any help
I use latest version of ES (6.2)