[
https://issues.apache.org/jira/browse/HDDS-14194?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ivan Andika resolved HDDS-14194.
--------------------------------
Resolution: Implemented
> Make streaming write SYNC configurable
> --------------------------------------
>
> Key: HDDS-14194
> URL: https://issues.apache.org/jira/browse/HDDS-14194
> Project: Apache Ozone
> Issue Type: Sub-task
> Reporter: Ivan Andika
> Assignee: Ivan Andika
> Priority: Major
> Labels: pull-request-available
>
> See BlockDataStreamOutput#needSync, currently SYNC is never used and
> therefore we might not be sure when the data is actually flushed to the disk
> (see StreamDataChannelBase#force which will invoke FileChannel#force(false)
> to flush the data to the underlying storage)
> {code:java}
> private boolean needSync(long position) {
> if (SYNC_SIZE > 0) {
> // TODO: or position >= fileLength
> if (position - syncPosition >= SYNC_SIZE) {
> syncPosition = position;
> return true;
> }
> }
> return false;
> } {code}
> We can create a configuration to allow SYNC, which is disabled by default to
> preserve compatibility.
> We also might want to revisit whether we should make SYNC to be a client-side
> or server-side. The concern is if a client is misconfigured, it might cause a
> lot of sync on the DN which will cause higher latencies.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
