Broker may lost messages in extreme condition

[JosephChanChan]

There a problem when mapfile flush message to disk.
details see: #3129

The measures bellow:

• message property: MessageConst.PROPERTY_WAIT_STORE_MSG_OK to true
• broker config : flushDiskType=FlushDiskType.SYNC_FLUSH

It can't ensure a message could be flush to disk. the flushedPosition would updated to writePosition even if a error was occurred. then return a PUT_OK result to producer in GroupCommitService.doCommit function

there some messages will be lost if power off now, and the producer has been received a ok response.

[labazhou2018]

In most case, the data in pageCache , it is safe.
When the message is in MappedByteBuffer or FileChannel, RocketMQ think the message is safe.
Java Program call flush(), then system flush data from pageCache to disk( it only depends on system)

[labazhou2018]

broker config : flushDiskType=FlushDiskType.SYNC_FLUSH，should use "this.mappedByteBuffer.force()"

MappedByteBuffer force():

[github-actions]

This issue is stale because it has been open for 365 days with no activity. It will be closed in 3 days if no further activity occurs.

[github-actions]

This issue was closed because it has been inactive for 3 days since being marked as stale.