Cluster-Related Advanced Properties


THE INFORMATION IN THIS ARTICLE APPLIES TO:

  • EFT v8.0 and later

DISCUSSION

The advanced properties described below affect EFT HA (active-active) configurations. These values are read upon EFT server service startup.

Specify the advanced properties described in the table in the format shown below. For more information about the Advanced Properties, refer to the online help for your version of EFT.

{

"ClusterOutOfSyncGracePeriodSecs": 10

"ClusterOutOfSyncHealSecs": 60

}

Additional cluster-related advanced properties:

Name Type Min Max Default Description
ClusterCoherenceQueueDetectPrivateIP string 0 128 autodetect Determine whether to automatically detect private IP to advertise to other nodes. Default value is "autodetect". You can replace "autodetect" with the prefix of the IP address to find (e.g: "192.168" to match an IP address starting with 192.168).
ClusterCoherenceQueueMsmqType string 0 128 msmq-multicast Determines the type of coherence queue to use when using MSMQ. msmq-multicast uses multicast-based MSMQ queues to send administrative updates. msmq-iterative uses point-to-point MSMQ queues (Unicast). Default value is msmq-multicast.
ClusterCoherenceQueueMulticastConfirmationTimeoutSecs uint32_t 0 2147483647 30 Specifies how much time (in seconds) EFT HA cluster node should wait for own multicast MSMQ message to arrive to its own coherence queue before failing send attempt.
ClusterCoherenceQueueMulticastMaxRetryCount uint32_t 0 2147483647 5 Specifies how much retry attempts EFT HA cluster mode should make when sending multicast MSMQ message to coherence queue.
ClusteredRuleHeartbeatPeriodSecs uint32_t 10 3600 10 This specifies the heartbeat period in seconds for clustered event rules. The smaller the period, the sooner nodes will notice when another node has gone down, but more frequent heartbeats incur increased overhead for the production and processing of the associated event rule queue traffic.
 ClusterOutOfSyncGracePeriodSecs  string 10 60 10

Amount of time in seconds that an HA node will wait for incoming administrative messages to arrive before declaring itself to be out-of-sync with the cluster.  

Enables advanced property below.

ClusterOutOfSyncHealSecs  uint32_t 0 (disables auto-draining) 4294967295   60 seconds

Amount of time in seconds that an HA node will wait for incoming administrative messages to arrive before declaring itself to be out-of-sync with the cluster and initiating draining and restart.

  1. Provide advanced property, ClusterOutOfSyncGracePeriodSecs, (above) and change the full timeout for sync default from 30 seconds to 60 seconds.
  2. Provide advanced property, ClusterOutOfSyncHealSecs, to set this value to non-default value.
  3. When EFT becomes out of sync and enters the period where it attempts to heal,  the WEL text is output: "Node out of sync, attempting to heal...".
  4. The EFT.log shows: 
  • Time it entered into out of sync state: "Initial detection of node as out of sync, entering grace period..."
  • Time it attempted to start healing itself: "Node out of sync, attempting to heal..."
    Time it took to heal (if healed): "Node was out of sync for [N] seconds but is now in sync"
  • Time at which it considered itself unable to heal; Time at which it began to drain: "Out of sync node unable to recover after [N] seconds, entering drain mode for maximum of [M] minutes. Service [will\will not] restart when draining is complete"
  • Time at which it completed drain: Elapsed time for completing drain "Finished draining node after [N.XY] minutes"