none
Log records for sql azure geo replication failures RRS feed

All replies

  • In geo-replication the transactions are sent to the secondary as soon as they are committed. They are replayed on the secondary as soon as they are received. It is not possible that tx 13 is lost but tx 14 is successful. If an error causes synchronization to keep failing it results in the replication log size increase on the primary. When the log size reaches certain threshold the secondary will be terminated and a new one will be created automatically. This is called re-seeding.

    The value you should be monitoring is last_replication in sys.dm_geo_replication_link_status (populated on the primary database). It allows you to estimate the potential data loss after failover.

    Thank you

    Thursday, June 23, 2016 9:34 PM
  • If the replication message fails we have retry logic to repeat. If it keeps failing we will hold all subsequent transactions and therefore the lag will increase. We monitor the lag internally and take actions if the lag gets too large. In the extreme case the service will automatically re-seed the entire database to recover from this condition. We don't have a log but the cuhstomer can monitor the DMV as described in the other answer.

    Would be good to understand what the customer would do if such log were exposed.

    Friday, July 8, 2016 5:42 PM