org.reveno.atp.clustering.api

Interface ClusterConfiguration.SyncConfiguration

All Known Implementing Classes:

RevenoClusterConfiguration.RevenoSyncConfiguration

Enclosing interface:

ClusterConfiguration

public static interface ClusterConfiguration.SyncConfiguration

Method Summary

Modifier and Type	Method and Description
void	mode(SyncMode mode) A mode in which two nodes will perform synchronization during Leadership Election.
void	port(int port) Port on which Sync Server will be listening.
void	retries(int count) Number of retries to transmit state data until it fails.
void	threadPoolSize(int threads) Thread pool size of Sync Server, which handles synchronization requests and sends appropriate data.
void	waitAllNodesSync(boolean wait) The synchronization process lasts during global Leadership Election process, which means no node during it can operate on production.

Method Detail

mode

void mode(SyncMode mode)

A mode in which two nodes will perform synchronization during Leadership Election. First option means that all nodes makes snapshots in advance and then just send it by request from other nodes. Second option is when the node sends required Journal files to other nodes, which contains events with transaction ID more or equals to current node last transaction ID. First option is prefered, but is not recommended for big models, because of size of snapshot and costs spent to make it for each Leadership Election process. Second option is generally not recommended if there is a big amount of Journal files or rolling is performed very rarely, means big files will be always transmitted. Default value is SyncMode.SNAPSHOT

Parameters:

mode -

threadPoolSize

void threadPoolSize(int threads)

Thread pool size of Sync Server, which handles synchronization requests and sends appropriate data.

Parameters:

threads - to be used

retries

void retries(int count)

Number of retries to transmit state data until it fails.

Parameters:

count -

port

void port(int port)

Port on which Sync Server will be listening.

Parameters:

port -

waitAllNodesSync

void waitAllNodesSync(boolean wait)

The synchronization process lasts during global Leadership Election process, which means no node during it can operate on production. Sometimes, it is obvious that Master candidate has the latest state, and can start to operate not awaiting other Slaves to catch up with its changes. In that case, if waitAllNodesSync is set to true, it will start to handle Commands normally during "sync" stage of Leadership Election process. Warning!!! You should consider using that flag *very* carefully, as it means that all failover data from Master will not be handled by Slaves as usual, but just buffered in memory, until they finish syncing. And if synchronization process takes too long, some data might be even lost. Use it primarily if you want Master to start operate as soon as possible (millisends matters) and you know that sync process either not happens at all or will be very short.

Parameters:

wait -