Leveraging the AWS Cloud to Create Scalable Research Performance Analytics
2 likes432 views
The document outlines the recovery processes for various components of a distributed computing system, specifically regarding the HPCC framework. It details how auto-scaling groups address instance failures by terminating and replacing impaired instances while maintaining stable network configurations with elastic network interfaces (ENIs). Recovery mechanisms are described for Thor and Roxie instances through replication and redundancy strategies, ensuring continued operation despite node losses.
Leveraging the AWS Cloud to Create Scalable Research Performance Analytics
- 3. • • •
- 15. • • • • • • •
- 16. • • • • • •
- 17. • • • •
- 18. • • • •
- 22. • Auto Scaling Group detects impaired instances, terminates the instance and brings up new one in its place. • New instance comes up with different IP, hence not recognized by HPCC configuration. • All Instances bound to Elastic Network Interfaces (ENI) to maintain stable secondary private IP. • Secondary Private IP are used in HPCC configuration by binding to secondary Ethernet interface (eth1) • Recovery from Thor Slave Instance loss • Thor has replication is turned on by default. • Thor Slave data recovered from adjacent instance by running “ • Recovery from Thor master Instance loss • Configured HPCC to turn on DALI replication • DALI metadata recovered from Dali replication instance.
- 23. • Recovery from Roxie slave Instance loss • Configured HPCC to turn on cyclic redundancy. • Data automatically recovered when the replacement instance comes back up. • Even with the loss of a node, Roxie continues to function
- 24. • • • • • • •
- 25. • • • • • •
- 27. • • • • • •
- 31. • • • • •
- 32. • • • • • • • •
- 37. • • • • •
- 38. • • • •
- 43. • • • • • • •