Asynchronous mode for HDR replication

Asynchronous HDR replication means that the primary server does not wait for a response from the HDR secondary server before flushes the logical log to disk. Asynchronous HDR replication can increase replication speed, but transactions can be lost.

There are multiple ways to turn on asynchronous mode for HDR replication:

Set the DRINTERVAL configuration parameter to a positive integer value.
Set the DRINTERVAL configuration parameter to 0, and set the HDR_TXN_SCOPE configuration parameter to ASYNC.
Run the following statement:
```
SET ENVIRONMENT HDR_TXN_SCOPE 'ASYNC';
```

In asynchronous mode, the primary database server flushes the logical-log to disk after it copies the contents of the logical-log buffer to the data-replication buffer. The primary database server sends the contents of the HDR buffer across the network when one of the following conditions occurs:

The HDR buffer becomes full.
The time interval since the records were sent to the HDR secondary database server exceeds the value of the primary server's DRINTERVAL configuration parameter.

To reduce the risk of lost transactions in a cluster that uses asynchronous replication, use unbuffered logging for all the databases. Unbuffered logging reduces the amount of time between transaction-record writing and transfer. If your primary server uses buffered logging, and you receive an error -7350 Attempt to update a stale version of a row message, switch to unbuffered logging.

If a failover does occur, but the primary server is restarted with data replication, transactions that were committed on the primary server and not committed on the secondary server are stored in a file that is specified by the DRLOSTFOUND configuration parameter.