Skip to content

June 2026

The Summer 2026 maintenance window coincides with the campus data center generator test and is scheduled for the week of June 15th. HPC@UCD will be bringing servers down on Monday, June 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 kernel
  • NVidia GPU drivers
  • Open OnDemand
  • Quobyte
  • Slurm
  • ZFS
  • Ethernet switch firmware
  • InfiniBand card firmware

Please note that accounts that are no longer in UC Davis IAM will go through the separation process. Please note that TAF accounts are in IAM.

The ability to encrypt data at rest will be enabled for Quobyte volumes.

Farm Specific

All job data written to /tmp and /scratch will be automatically deleted when your Slurm job exits. These spaces are still a shared resource, so please be a good community member and not overfill them.

Farm sells hardware and storage with a five-year guarantee. Many Farm hardware components have exceeded their lifecycle but will NOT be retired during the June 2026 maintenance window. Please note that if your data storage is older than five years, you will get a warning upon login. Farm does NOT back up data, and older storage is in danger of data loss. Please act accordingly.

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.

Franklin Specific

None at this time.

Hive Specific

All job data written to /tmp and /scratch will be automatically deleted when your Slurm job exits. These spaces are still a shared resource, so please be a good community member and not overfill them.

The /scratch/nfs convenience mount no longer exists. If you need to access the network scratch, you can use the full path /nfs/hive/scratch. This area is no longer automatically deleted at the end of every job, so you can use it for temporary storage for data that needs to be accessed by multiple jobs. It is your responsibility to remove this data when you no longer need it. Abuse of this storage can result in your jobs being killed and your account being locked.