Hydra will remain powered up for use during the government shutdown.
The data center will have essential personnel through the shutdown
to maintain the integrity of the IT infrastructure.
Access to Hydra will not be cut off, so users will be able to submit jobs.
Any major failure of the IT infrastructure (hardware or software) is unlikely
to be fixed during the shutdown, hence accessibility to Hydra and its resources
cannot be guaranteed.
Support for Hydra will be limited to urgent matters
please contact SI-HPC-Admin@si.edu during the shutdown.
Hydra has been successfully moved to the new data center and is operational.
Updates are described in detail at the
HPC Wiki 2025 Data Center Move page.
Please take the time needed to read these pages before contacting us for support.
You can view the list of all the available modules:
as an HTML document, or
a plain ASCII text file.
You can also check the bandwidth between
SAO and HDC.
You can select to have this page refreshed every 5m,
20m, or
1hr, this one will auto-refresh every
1hr.
Usage
Current snapshot sorted by nodes'
.
Usage vs time, for length=
and user=
highlighted.
As of Sun Nov 9 20:37:03 2025: #CPUs/nodes 5740/74, 0 down. Loads:
head node: 0.78, login nodes: 0.15, 0.01, 24.07, 0.07; NSDs: 0.69, 0.29, 1.54, 5.63, 6.78; licenses: none used.
Queues status: 9 disabled, none need attention, none in error state.
14 users with running jobs (slots/jobs):
Total number of queued jobs/tasks/slots: 3/1,790/4,295
55 users have/had running or queued jobs over the past 7 days, 65 over the past 15 days.
83 over the past 30 days.
Click on the tabs to view each section, on the plots to view larger versions.
You can view the current cluster snapshot sorted by name, no. cpu, usage, load
or memory, and
view the past load for 7, or 15, or 30 days as well as highlight a given user by
selecting the corresponding options in the drop down menus.
{}
This page was last updated on Sunday, 09-Nov-2025 20:41:56 EST
with mk-webpage.pl ver. 7.3/1 (Oct 2025/SGK) in 0:37.
Warnings
Oversubscribed Jobs
As of Sun Nov 9 20:37:05 EST 2025 (1 oversubscribed job)
Total running (PEs/jobs) = 1205/493, 3 queued (jobs), showing only oversubscribed jobs (cpu% > 133% & age > 1h) for all users.
jobID name user age nPEs cpu% queue node taskID
11030577 job_10a_auto_bi scottjj 05:07 4 463.7% mThM.q 65-03
⇒ Equivalent to 14.5 overused CPUs: 4 CPUs used at 463.7% on average.
Inefficient Jobs
As of Sun Nov 9 20:37:06 EST 2025 (1 inefficient job)
Total running (PEs/jobs) = 1204/492, 3 queued (jobs), showing only inefficient jobs (cpu% < 33% & age > 1h) for all users.
jobID name user age nPEs cpu% queue node taskID
11030567 phyluce_spades guerravc 07:21 16 29.0% sThC.q 65-27
⇒ Equivalent to 11.4 underused CPUs: 16 CPUs used at 29.0% on average.
Nodes with Excess Load
As of Sun Nov 9 20:37:08 EST 2025 (5 nodes have a high load, offset=1.5)
#slots excess
node #CPUs used load load
-----------------------------------
65-03 64 8 43.4 35.4 * 75-02 128 26 29.0 3.0 * 76-03 192 20 30.1 10.1 * 84-01 112 14 16.3 2.3 * 93-06 96 16 27.6 11.6 *Total excess load = 62.4
As of Sun Nov 9 20:37:07 EST 2025
2 jobs waiting for breusingc:
jobID jobName user age nPEs memReqd queue taskID
--------- --------------- ---------------- ------ ---- -------- ------ -------
11024487 metaspades breusingc +9:11 16 0.0 mThM.q 341-402:1
11026977 AssembleBarcode breusingc +5:11 16 256.0 sThM.q 298-402:1
quota rule resource=value/limit %used
------------------- ------------------------------- ------
max_slots_per_user/1 slots=480/840 57.1% for breusingc
max_hM_slots_per_user/2 slots=320/585 54.7% for breusingc in queue mThM.q
max_mem_res_per_user/2 mem_res=2.500T/8.944T 28.0% for breusingc in queue uThM.q
max_hM_slots_per_user/1 slots=160/840 19.0% for breusingc in queue sThM.q
------------------- ------------------------------- ------
1 job waiting for jbak:
jobID jobName user age nPEs memReqd queue taskID
--------- --------------- ---------------- ------ ---- -------- ------ -------
11030566 OMI-2025m0616t jbak 07:22 1 mThC.q 1748-3369:1
quota rule resource=value/limit %used
------------------- ------------------------------- ------
max_slots_per_user/1 slots=438/840 52.1% for jbak
max_hC_slots_per_user/2 slots=438/840 52.1% for jbak in queue mThC.q
max_mem_res_per_user/1 mem_res=876.0G/9.985T 8.6% for jbak in queue uThC.q
------------------- ------------------------------- ------
Overall Quota Usage
quota rule resource=value/limit %used
------------------- ------------------------------- ------
total_slots/1 slots=1205/5960 20.2% for *
total_mem_res/3 mem_res=1.465T/7.874T 18.6% for * in queue uTxlM.rq
total_mem_res/2 mem_res=5.605T/35.78T 15.7% for * in queue uThM.q
blast2GO/1 slots=14/110 12.7% for *
total_mem_res/1 mem_res=1.072T/39.94T 2.7% for * in queue uThC.q
Memory Usage
Reserved Memory, All High-Memory Queues
Select length:
Current Memory Quota Usage
As of Sun Nov 9 20:37:08 EST 2025
quota rule resource=value/limit %used filter
---------------------------------------------------------------------------------------------------
total_mem_res/1 mem_res=1.072T/39.94T 2.7% for * in queue uThC.q
total_mem_res/2 mem_res=5.605T/35.78T 15.7% for * in queue uThM.q
total_mem_res/3 mem_res=1.465T/7.874T 18.6% for * in queue uTxlM.rq
Current Memory Usage by Compute Node, High Memory Nodes Only
Limit slots for all users together
users * to slots=5960
users * queues sThC.q,lThC.q,mThC.q,uThC.q to slots=5176
users * queues sThM.q,mThM.q,lThM.q,uThM.q to slots=4680
users * queues uTxlM.rq to slots=536
users * queues sTgpu.q,mTgpu.q,lTgpu.q to slots=104
Limit slots/user for xlMem restricted queue
users {*} queues {uTxlM.rq} to slots=536
Limit total reserved memory for all users per queue type
users * queues sThC.q,mThC.q,lThC.q,uThC.q to mem_res=40902G
users * queues sThM.q,mThM.q,lThM.q,uThM.q to mem_res=36637G
users * queues uTxlM.rq to mem_res=8063G
Limit slots/user for interactive (qrsh) queues
users {*} queues {qrsh.iq} to slots=16
Limit GPUs for all users in GPU queues to the avail no of GPUs
users * queues {sTgpu.q,mTgpu.q,lTgpu.q,qgpu.iq} to GPUS=8
Limit GPUs per user in all the GPU queues
users {*} queues sTgpu.q,mTgpu.q,lTgpu.q,qgpu.iq to GPUS=4
Limit GPUs per user in each GPU queues
users {*} queues {sTgpu.q} to GPUS=4
users {*} queues {mTgpu.q} to GPUS=3
users {*} queues {lTgpu.q} to GPUS=2
users {*} queues {qgpu.iq} to GPUS=1
Limit to set aside a slot for blast2GO
users * queues !lTb2g.q hosts {@b2g-hosts} to slots=110
users * queues lTb2g.q hosts {@b2g-hosts} to slots=1
users {*} queues lTb2g.q hosts {@b2g-hosts} to slots=1
Limit total bigtmp concurrent request per user
users {*} to big_tmp=25
Limit total number of idl licenses per user
users {*} to idlrt_license=102
Limit slots for io queue per user
users {*} queues {lTIO.sq} to slots=8
Limit slots for io queue per user
users {*} queues {lTWFM.sq} to slots=2
Limit the number of concurrent jobs per user for some queues
users {*} queues {uTxlM.rq} to no_concurrent_jobs=3
users {*} queues {lTIO.sq} to no_concurrent_jobs=2
users {*} queues {lWFM.sq} to no_concurrent_jobs=1
users {*} queues {qrsh.iq} to no_concurrent_jobs=4
users {*} queues {qgpu.iq} to no_concurrent_jobs=1
Limit slots/user in hiCPU queues
users {*} queues {sThC.q} to slots=840
users {*} queues {mThC.q} to slots=840
users {*} queues {lThC.q} to slots=431
users {*} queues {uThC.q} to slots=143
Limit slots/user for hiMem queues
users {*} queues {sThM.q} to slots=840
users {*} queues {mThM.q} to slots=585
users {*} queues {lThM.q} to slots=390
users {*} queues {uThM.q} to slots=73
Limit reserved memory per user for specific queues
users {*} queues sThC.q,mThC.q,lThC.q,uThC.q to mem_res=10225G
users {*} queues sThM.q,mThM.q,lThM.q,uThM.q to mem_res=9159G
users {*} queues uTxlM.rq to mem_res=8063G
Limit slots/user for all queues
users {*} to slots=840
You can view plots of disk use vs time, for the past 7, 30, or 120 days;
as well as
plots of disk usage by user, or
by device (for the past 90 or 240 days respectively).
Notes
Capacity shows % disk space full and % of inodes used.
When too many small files are written on a disk, the file system can become full because it is
unable to keep track of new files.
The % of inodes should be lower or comparable to the % of disk space used.
If it is much larger, the disk can become unusable before it gets full.