VOMS server resilience

1
VOMS server resilience
22nd GridPP Collaboration Meeting

• Voms.gridpp.ac.uk
• S.Dolgobrodov
• Manchester Tier2

2
Main vulnerable components

• Computer hardware
• Main power
• Network access
• Database
• Hosting about 20 regional Vos, more than 300
  users in total

3
Hardware resilience

• Machine hardware
• Backup machine with alias NI device and IP
  address
• This is /etc/sysconfig/network-scripts/ifcfg-e
  th01
• DEVICEeth01
• IPADDR195.194.110.230
• 195.194.110.231 voms01.gpp.hep.man.ac.uk
  
• 195.194.110.232 voms02.gpp.hep.man.ac.uk
  
• 195.194.110.230 voms.gridpp.ac.uk
• Power supply
• Separate power lines for the front and backup
  servers
• Emergency power backup generator for the farm,
  air conditioners maintained round the clock.
• Loss of network
• Separate switch for the front and back up
  servers.

4
Software resilience etc

• Data base backup
• MySQL data bases (one per each VO) daily backing
  up on the backSecoup servers (3 replicas) keeping
  them mirrored.
• The only records under risk are those created
  during the day.
• Currently more than 300 records in 19 VOs
• Security
• ssh access restricted based on IP addresses
• ssh host keys are not used for access
• distinctive password from other farm nodes
• Human factor
• A second administrator with documented
  procedures for recovery.