The document discusses high availability and fault tolerance using Novell Cluster Services. It defines key concepts like availability, mean time between failures, and mean time to repair. It then covers best practices for deploying Novell Cluster Services, including hardware and software setup, connectivity rules, naming and addressing, and testing the cluster. It also discusses which types of resources can be clustered, like file sharing, iPrint, iFolder, and DHCP.

1. High-Availability with Novell Cluster Services ™ for Novell ® Open Enterprise Server on Linux Tim Heywood , CTO, NDS8 [email_address] Martin Weiss , Senior Technical Specialist [email_address] Dr. Frieder Schmidt , Senior Technical Specialist [email_address]

2. Agenda High Availability and Fault Tolerance Novell Cluster Services ™ Best Practices Deploying Cluster Services What is Clusterable? Demo

3. High-Availability and Fault Tolerance

4. High-Availability: Motivation Murphy's Law is universal: faults will occur Power failures, hardware crashes, software errors, human mistakes... Unmasked faults show through to the user

5. How much does downtime of a service cost you? Even if you can afford a 5 second blip, can you afford a day long outage or worse, loss of data? Can you afford low availability systems? If you are selling or depending on a service, service unavailability translates to cost

6. Definition: Availability Mean Time Between Failures (MTBF) follows a normal distribution Mean Time To Repair (MTTR)

7. Availability Percentage of time that a system functions as expected

8. Always computed for a certain time, i. e. a month, a year Example: MTBF: 360 days

9. MTTR: 1 hour

10. How to Determine Availability? Availability of a complex system is determined by the availability of its individual components

11. two ways to couple components: serial design

12. parallel design Availability of a serial design: A ser = A 1 * A 2 ; A 1 = 0.99, A 2 = 0.99, A ser = 0.9801

13. Availability of a parallel design: A par = 1 – ( 1 - A 1 ) * ( 1 – A 2 ); A par = 1 – ( 1 - 0.99 ) * ( 1 – 0.99 ); A par = 1 – ( 0.01 ) * ( 0.01 ) = 0.9999

14. “3R Rule” for High-Availability Systems R edundancy, R edundancy, R edundancy Fault Tolerance “The ability of a system to respond gracefully to an unexpected hardware or software failure.” Webopedia Computer System Fault Tolerance “The ability of a computer system to continue to operate correctly even though one or more of its components are malfunctioning.” Institute for Telecommunication Services, National Telecommunications and Information Administration, US Dept. of Commerce

15. Managing Risk: Two Goals Primary Goal: Increase Mean Time to Failure (MTTF) Choose reliable hardware

16. Implement redundant / fault tolerant systems Easy to implement for some components (power supplies, LAN connectivity, SAN connectivity, RAID, etc.)

17. Not so easy for other components (main board, memory, processor, etc. Establish sound administrative practices Secondary Goal: Reduce Mean Time to Repair (MTTR) Keep hardware spares close at hand

18. Document repair procedures and train personnel

19. Chose Open Enterprise Server– Linux Server with Novell Cluster Services ™

20. High-Availability by Clustering Redundant setup “clustered” to act as one avoid Single Point of Failure (SPOF) Primary focus is availability , but can allow for increased performance HA via fail-over: In case [an application on] a server failure is detected, another server takes over Results achieved depend on failure detection time and startup delays The [virtual] hand moves faster than the eye The fault is masked before the user really notices

21. Depends on failure detection time, restart time, overhead

22. Novell Cluster Services ™

23. Novell Cluster Services ™ Cluster services allows a resource to be activated on any host in the cluster

24. Load distribution over multiple servers when having multiple resources

25. Monitors LAN and SAN/Storage connectivity – in the event of a failure – fences the problematic node and relocates the resource

26. Supports active-passive clustering

27. Supports resource monitoring

28. Supports Linux and Novell ® Open Enterprise Server services

29. Supports up to 32 nodes per cluster

30. Novell Cluster Services ™ Easy Management

31. Easy Configuration Load Script

32. Unload Script

33. Monitoring Script iManager integration

34. Command Line Interface

35. E-mail and SNMP Notification

36. Integration with Novell ® Open Enterprise Server Services

37. Integration with XEN

38. Novell Cluster Services ™ Ctrl 2 Dual NICs Dual HBAs LUN 0 LUN 1 LUN … Ctrl 1 LAN Fabric SAN Fabric Storage Array Storage Array Novell iSCSI Storage Array Typical NCS 1.8 Architecture Fibre Channel or iSCSIl Ethernet

39. Cluster Services in Novell ® Open Enterprise Server (OES) 2 New features are Linux only

40. New from OES2 FCS on: Resource monitoring

41. XEN virtualization support

42. x86_64 platform support Including mixed 32/64 bit node support Dynamic Storage Technology

43. What's New in SP1/2? Major rewrite of cluster code for SP2 Removed NetWare ® translation layer

44. Much faster

45. Much lower system load

46. Typical load average of 0.2! New/improved clustering for: iFolder 3

47. AFP

48. …

49. … NCP ™ virtual server for POSIX filesystem resources :-(