TimesTen内存数据库是否支持高可用？若支持该如何配置？

Absolutely! TimesTen In-Memory Database fully supports high availability—its Active Standby Pair configuration is built specifically to keep your mission-critical workloads running without interruption, even if a primary node goes down. Let me break down what you need to know, plus a step-by-step setup guide to get you started.

Yes, absolutely. The core high availability feature is the Active Standby Pair:

• The active (primary) node handles all read/write operations.
• The standby node maintains real-time synchronization with the primary via transaction logs.
• If the primary node fails, the standby can be quickly promoted to active (either manually or automatically) to minimize downtime.
• Once the original primary recovers, it can be set back to standby to restore the HA pair.

Prerequisites First #

Before you start, make sure you have:

• Two servers/nodes with identical versions of TimesTen installed (same patch level too—critical for compatibility).
• Unrestricted network connectivity between nodes (default replication port is 53396; ensure firewalls don't block it).
• The same OS user account running TimesTen on both nodes (to align file permissions).
• Matching memory configuration parameters (PermSize, TempSize) on both nodes—mismatches can break replication.

Configure DSNs on Both Nodes #

Edit your ODBC configuration file (either sys.odbc.ini for system-wide access or ~/.odbc.ini for user-specific) on each node:

• Primary Node DSN Example:

  [TT_PRIMARY]
  DataStore=/opt/timesten/ttstore/primary
  PermSize=4096
  TempSize=2048
  LogAutoTruncate=0
  LogBufMB=128
  LogFileSize=2048
  ConnectionCharacterSet=AL32UTF8
  ReplicationPort=53396
  

• Standby Node DSN Example:

  [TT_STANDBY]
  DataStore=/opt/timesten/ttstore/standby
  PermSize=4096
  TempSize=2048
  LogAutoTruncate=0
  LogBufMB=128
  LogFileSize=2048
  ConnectionCharacterSet=AL32UTF8
  ReplicationPort=53396
  

Note: LogAutoTruncate=0 is required to prevent logs from being deleted before they're replicated to the standby.

Create the Active Standby Pair #

Connect to the primary node's DSN using ttIsql:

ttIsql TT_PRIMARY

Then run the SQL command to define the replication pair (replace with your actual hostnames):

CREATE ACTIVE STANDBY PAIR TT_PRIMARY ON 'primary-node.yourdomain.com', TT_STANDBY ON 'standby-node.yourdomain.com';

Optional: If you need read-only replicas to scale read workloads, add them with the READ ONLY SUBSCRIBERS clause.

Start Replication #

• On the primary node (still in ttIsql):

  START REPLICATION;
  

• On the standby node, connect to its DSN with ttIsql TT_STANDBY and run:

  CALL ttRepStart;
  

Verify Replication is Working #

Test data synchronization to confirm everything’s set up correctly:

1. On the primary node, create a test table and insert data:

   CREATE TABLE ha_test (id INT PRIMARY KEY, data VARCHAR(100));
   INSERT INTO ha_test VALUES (1, 'TimesTen HA Test');
   

2. On the standby node, query the table:

   SELECT * FROM ha_test;
   

You should see the inserted row immediately if replication is functioning as expected.

Handle Failover & Failback #

• Manual Failover: If the primary node fails, promote the standby to active on the standby node:

  CALL ttRepStateSet('ACTIVE');
  

• Failback: Once the original primary is restored, set it back to standby:

  CALL ttRepStateSet('STANDBY');
  

• Automatic Failover: For hands-off failover, configure it using the ttRepAdmin tool with the -automaticTakeover option. This requires setting up a watchdog process to monitor node health.

• Always keep both nodes on the same TimesTen version and patch level—upgrade the standby first, then failover, then upgrade the original primary.
• Monitor replication status regularly with ttRepAdmin -showstatus TT_PRIMARY to catch issues early.
• Ensure sufficient disk space for transaction logs on the primary node—logs won’t truncate until they’re replicated to the standby.

内容的提问来源于stack exchange，提问作者user7880058