IBM BS029ML Self Help Guide - Page 64

Figure 2-8 gives an overview of the DB2 HADR. Ethernet (Service Network) DB01 pSeries 570 4x 1.65 Ghz AIX 5.3.0.0 + HACMP 5.3 8 Gb RAM en0 DB2 UDB ESE V8.2 FP 10 en2 (primary) en1 en4 en3 Gigabit Ethernet (for HADR Log transfer) DB02 en0 en2 en1 en4 en3 pSeries 550 2x 1.65 Ghz AIX 5.3.0.0 + HACMP 5.3 4 Gb RAM DB2 UDB ESE V8.2 FP 10 (standby) hba0 hba1 hba2 hba2 hba1 hba0 tty1 tty1 RS232C (for HACMP) Tape Library SAN Storage Area for Node 1 Data Logs SAN Storage Area for Node 2 Data Logs Tape Library Figure 2-8 DB2 HADR Without HADR, the length of time it takes to cut over from a database failure is unpredictable. It can take several minutes or hours before the failure is solved and the database is available. HADR enables failover and fallback between the two systems. The Standby database can take over as the Primary database with full DB2 functionality. After the failed old Primary is repaired, it can rejoin the HADR pair as a Standby database if the two copies of the database can be made consistent. After the original Primary database is reintegrated into the HADR pair as the Standby database, a failback operation can be performed so that the original Primary database is once again the Primary database. HADR requires the same hardware, OS, and DB2 software on the two systems (except for some minor differences). A high availability mechanism is still a requirement, as DB2 HADR does not have a fault tolerant detection feature. In addition, if the Log transfer network is down, HADR takeover cannot be done. So this network is very important. In this configuration, a dedicated Gigabit Ethernet segment is used in conjunction with Network Interface Backup (NIB) for redundancy. Note that an outage at the Log transfer network would cause the Primary to drop out of communication with the Standby (and if in Peer state, would cause the Primary to drop out of Peer state and run independent of the Standby). Once the network is repaired, the Primary and Standby would be able to eventually come back into Peer. Thus, a network outage at the Log transfer network would not result in a failure to process transactions, as seen by clients. An alternative to HACMP is Tivoli System Automation (TSA). TSA is now bundled with DB2 for AIX (as of DB2 ESE 8.2 FP 13) in the same manner, and same licensing terms, as Linux (TSA bundled with DB2 ESE 8.2 on Linux). All HACMP or TSA has to do is detect a node 50 IBM WebSphere Portal V6 Self Help Guide