December 2025
The fall 2025 maintenance window coincides with the campus data center generator test and is scheduled for the week of December 15th. HPC@UCD will be bringing servers down on Monday, December 15th, and they will be made available for use after planned maintenance tasks are completed. During this time, Farm, Franklin, and Hive will be unavailable for use, and all jobs will be stopped.
Jobs that have a runtime longer than the current time until the start of the maintenance window will not start. This
exact amount of time is printed every time you log in to each of the clusters. If your job is held because of the
maintenance reservation, you will see this in your squeue output:
(ReqNodeNotAvail, Reserved for maintenance)
All Clusters¶
During the maintenance window, the following software items will be updated:
- CVMFS (Software Distribution)
- Linux cgroups
- NVidia GPU drivers
- Open OnDemand
- Quobyte
- Slurm
- ZFS
Farm Specific¶
Farm sells hardware and storage with a five-year guarantee. Many Farm hardware components have exceeded their lifecycle and will be retired during the December 2025 maintenance window. At the request of CA&ES IT, all hardware more than five years old will be retired, and those Slurm partitions will be removed. Additionally, the partition names for the newer hardware will be renamed.
Farm Slurm Partition Status as of December 15, 2025¶
| Old partition name | New Name/Status |
|---|---|
high2 |
high |
med2 |
Deleted |
low2 |
low |
high |
Deleted |
med |
Deleted |
low |
Deleted |
bigmemh |
Deleted |
bigmemm |
Deleted |
bigmeml |
Deleted |
bit150h |
Deleted |
ecl243 |
Deleted |
bmh / bml |
Untouched |
bmm |
Deleted |
gpul |
Functionality folded into low. Access with: --partition=low --gres=gpu:TYPE:COUNT |
If you need to start an order for new hardware, see https://hippo.ucdavis.edu/Farm/product/index. Note that there are significant delays in the delivery of computing hardware.
New Public Access Group Name¶
The long-term group name used for free-tier Farm access will be changing. The new group name will be publicgrp, but
the description in Hippo will remain the same (CA&ES Free Tier sponsored by Adam Getchell). All users in the previous
group will be automatically transitioned.
New Login Node¶
Farm's login node will be upgraded to a new system. Changes:
- CPUs: 32 -> 256
- RAM: 64 GB -> 512 GB
Franklin Specific¶
None at this time.
Hive Specific¶
New Login Nodes¶
The two current virtual login nodes will be replaced with physical hardware. Changes:
- CPUs: 32 -> 64
- RAM: 64 GB -> 256 GB
Additionally, the physical machines have InfiniBand, which should reduce overhead when accessing storage.