cluster – FASRC DOCS

SEAS Compute Resources

The Harvard John A. Paulson School of Engineering and Applied Sciences has a number of compute resources managed by FAS Research Computing. Members of the SEAS community may access compute resources using this information. Available partitions and their component parts:

Partition Name	Assigned compute nodes	Description of nodes
holyseasgpu	holyseasgpu[01-07,09-11]	each node: Dell R720, 2 x Intel E5-2695 v2 @ 2.40GHz 12 core, 90GB RAM, 2 x Nvidia K40m
seas_gpu	holygpu2c0913	single node: Dell DSS 8440 Cauldron, 2 x Intel Xeon Gold 6148 2.4G 20 core, 373GB RAM, 10 x NVIDIA Tesla V100 32G, EDR Infiniband
seas_dgx1	seasdgx[101-104]	each node: Nvidia DGX1, 2 x Intel(R) Xeon(R) CPU E5-2698 v4 2.20GHz, 499GB RAM, 8 x NVIDIA Tesla V100 16G, FDR Infiniband
seas_dgx1_priority	see above	see above
seas_gpu_requeue	holyseasgpu[01-7,09-11], holygpu2c09[13,17,31], seasdgx10[1-4]	see individual entries
fdoyle_compute	seasmicro[01-05,07-28]	each node: Supermicro SYS-6018R-MTR, 2 x Intel(R) Xeon(R) CPU E5-2620 v3 2.40GHz 6 core, 247GB RAM, 1GbE
seas_iacs	holyseas0[5-6]	each node: Dell M915, 4 x AMD Opteron(tm) Processor 6376 16 core, 247GB RAM, 1GbE
narang_dgx1	seasdgx[101-102]	see individual entries
pierce	holyseas04,holy2b09303	both nodes Dell M915
mazur	holyvulis01	Dell M630, 2 x Intel Xeon CPU E5-2697 v4 2.30GHz 18 core, 247GB RAM, 1GbE, FDR Infiniband
tambe_gpu	holygpu2c0917	single node: Dell DSS 8440 Cauldron, 2 x Intel Xeon Gold 6148 2.4G 20 core, 384GB RAM, 10 x NVIDIA Tesla V100 32G, EDR Infiniband
barak_ysinger_gpu	holygpu2c0931	single node: Supermicro SYS-4029GP-TRT, 2 x Intel Xeon Silver 4216 CPU 2.10GHz 16 core, 384GB RAM, 8 x NVIDIA GeForce RTX 2080, 1GbE, FDR Infiniband
idreos_parkes	holygpu2c1125	single node: Dell DSS 8440 Cauldron, 2 x Intel Xeon Gold 6248 2.5G 20 core, 1.5B RAM, 10 x NVIDIA Tesla V100 32G, EDR Infiniband
seas	holy7c026[09-12], holy7c241[05-12],holy7c242[05-12],holy7c243[05-12], holy7c244[05-12],holy7c245[05-12],holy7c246[05-12]	each node: Lenovo SD650v1, 2 x Intel Xeon Platinum 8268 @ 2.9 GHz 24 core, 192 GB RAM, EDR IB
seas_priority	holy7c026[09-12], holy7c241[05-12],holy7c242[05-12],holy7c243[05-12], holy7c244[05-12],holy7c245[05-12],holy7c246[05-12]	each node: Lenovo SD650v1, 2 x Intel Xeon Platinum 8268 @ 2.9 GHz 24 core, 192 GB RAM, EDR IB
Partition Name	Assigned Compute Nodes	Description of nodes

The partitions and the groups that have access to them are listed here. For restricted slurm compute partitions you will need to be in the proper AD group to submit jobs (type “id” on the command line to see the groups you are in). These groups have priority access to the listed resources; they are part of the general pool when not in use. Nearly all compute nodes are part of the “serial_requeue” or “gpu_requeue”. This is the backfill scheduler – any higher priority jobs will kick those off of the node if the resources are needed. Lowest priority and shortest running jobs are preempted first.

Partition Name	Group(s) Who Have Access	Additional notes
holyseasgpu	computefestgpu, gershman_lab, seas
seas_gpu	seas
seas_dgx1	seas
seas_dgx1_priority	acc_lab rush_lab kung_lab
seas_gpu_requeue	seas
fdoyle_compute	fdoyle_lab
seas_iacs	seas_iacs
narang_dgx1	narang_lab
pierce	pierce_lab slurm_group_pierce
mazur	mazur_lab_seas
tambe_gpu	tambe_lab
barak_ysinger_gpu	barak_lab ysinger_group
idreos_parkes	idreos_lab parkes_lab
seas	slurm_group_seas	Is requeued by higher priority jobs in seas_priority
seas_priority	slurm_group_seas	Only groups with a higher fairshare than 0.75 may submit to this partition
Partition Name	Group(s) Who Have Access

For more information about Slurm partitions on the FAS RC cluster, please refer to the Running Jobs document.

Adding additional lab groups or cluster access

GROUP MEMBERSHIP

If you already have a FAS RC account (Note: multiple accounts are not allowed) and need to request membership in an additional lab group, the following will explain how to do so.

Your account can belong to multiple groups, with your PI/sponsor’s lab group being your primary. If you need to submit jobs under different labs groups you belong to, see: How do I fairly manage dual/multiple lab affiliations for work on the FASRC cluster?

CHANGING LABS

If you need to change your primary PI/sponsor to another lab group, please see How do I change lab groups?.

ADDING GROUPS

Requesting additional groups for file access can be done via our Portal.

Log into the Portal and select ‘Cluster Access’ on the left (or click here: https://portal.rc.fas.harvard.edu/request/ )
Two new links should appear on the left
– My Grants – Shows you all your current grants/memberships
– Add Grants – Lets you request new grants/memberships
Find the grant you wish to add and select its check box. A window will pop up allowing you to Cancel or Request Selected.

After clicking Request Selected, an approval message will be sent to the group’s approver (usually the PI, their lab manager, admin, etc.)

Once they approve your request, you will receive a notification via email.

You can check the status of your request or the groups you currently have by clicking on the My Grants link at any time.

NOTE: If you will need to be able to run jobs under that secondary group’s fairshare, please send FASRC a ticket (via Portal or Email) asking to be added to the lab’s SLURM group. See managing dual/multiple lab affiliations for details on using switching fairshare affiliations on jobs.

ADDING CLUSTER ACCESS

If you did not initially request cluster login access, you can also use the procedure above to request Cluster Access. This will allow you to log into the cluster and submit jobs. Go to Add Grants, Click ‘Cluster’ (if it does not appear, then you already have cluster access) to select the ‘FASRC cluster access’ checkbox.

If you are trying to SSH into the cluster and are constantly asked for your password or get an error about your home directory, this is the cause.

SPINAL

For SPINAL access questions, please see: Instrument SignUp/SPINAL