...
- SCS is planning a strategic change to GPFS and is now experimenting with the two new Fermi servers. One possible outcome of this work is to retain GPFS and run xroot on top of it. If not, then the machines will be reinitialized, the disks reformatted to XFS and configured as earlier Dell super-servers.
- There is worry about mixing xroot and NFS on the same server so that practice has been avoided in the past.
...
Option 1:
- Dedicate the two new servers to xroot service.
- NFS upgrade
- Migrate all xroot data from fermi-xrd001 (90 TB, installed 3/6/2012) to other servers
- Reconfigure fermi-xrd001 for NFS use
- Move contents of wain026, 031 and 032 to fermi-xrd001
- Move user partitions on wain025 to wain031 (6-months newer than wain025)
- Move group partitions on wain025 to wain032
- Retire wain025 and wain026
- Select eight (8) additional wains in xroot service to vacate and retire.
...
- This idea might not be popular with ISOC as it bundles their currently independent file server, wain031, with other user groups on fermi-xrd001. They seem comfortable with the status quo so the remaining issue is the age of wain031 (installed 3/9/2009). (QuestonQuestion: Could one use VMs, CHOS, cgroups or some other mechanism to effectively partition a machines resources and then run separate instances of NFS on each partition to achieve some guarantee of performance for each NFS group?)
- While this option effectively doubles the performance for the user/group disks, as well as substantially increasing the amount of space available, it does not address the age of those servers (both installed March 2009).