D0 Calibration Servers Deployment on Linux

1
D0 Calibration Servers Deployment on Linux

• By Anil Kumar
• CD/CSS/DSG
• 14th April, 2004

2
Stake Holders

• DO
• CD

3
Project Scope

• Putting the servers into kits.
• Working on bugs, and testing and finally
  implement on Linux node. This is first
  implementation of D0 Db servers on Linux OS.
• Provide a capability to Support the D0 Servers on
  Linux in 247 mode by implementing automatic
  failover solution.
• Dedicated failover solution for each user and
  farm servers with e-mail notification to admin
  and support groups.

4
Project Scope Contd -

• Provide a capability to view the server logs on
  web for troubleshooting.
• Train the SAM shifters to support the servers.
  Write Training instructions.
• Web Site for the project for view logs,
  diagnostic wage and failover plan and other
  related info.
• Procedure to stop and start one server out of set
  of servers.

5
Deployment Strategy

• Linux servers for the reconstruction farm and
  Analysis farm (user clients).
• A failover server box which will provide the
  required 24/7 mode of services.

6
Deployment Infrastructure
Linux PC for Analysis Farm
Linux PC for Reconstructed Farm
Sun Solaris
Linux PC
Linux PC
Failover
Failover
7
Status

• Project completed successfully as defined.
• Deployed the D0 Calib User and farm Servers on
  Linux Nodes. UPS/UPD the servers. Primary
  Testing before deploying.
• Implemented a automatic failover (for 247
  Support) to Secondary node if Primary node fails
  with e-mail notification to admin group and
  responsible parties.
• Capability to view the server logs on Web.
• Training to Sam Shifters to support Servers in
  247 Mode
• Instructions document for support of servers.
• Web site for D0 calib Servers Project
• http//d0db-prd.fnal.gov/d0dbsrv
• CVS d0dbsrv module.

8
Status Contd -

• April 2002 Deployment on d0dbsrv1 with failover
  on d0dbsrv2
• Feb 2003 - Migration of Servers to Reliable ASA
  machines. Each node serving a special set of
  servers with failover to single node.
• Feb 2004 Dedicated Failover node for user and
  farm servers.