Cluster mode is one of three primary ways of running Space Ranger. To learn about the other approaches, see the Different Ways of Running Space Ranger page.
Space Ranger can be run in cluster mode using Slurm to run the stages on multiple nodes via batch scheduling. This allows highly parallelizable stages to use hundreds or thousands of cores concurrently, dramatically reducing time to solution.
Running pipelines in cluster mode requires the following:
- Space Ranger is installed in the same location on all nodes of the cluster. For example,
/opt/spaceranger-2.1.0
or/net/apps/spaceranger-2.1.0
. - Space Ranger pipelines are run on a shared file system accessible to all nodes of the cluster. NFS-mounted directories are the most common solution for this requirement.
- The cluster accepts both single-core and multithreaded (shared-memory) jobs.
Installing the Space Ranger software on a cluster is identical to installation on a local server. After confirming that the spaceranger
commands can run in single server mode, configure the job submission template that Space Ranger uses to submit jobs to the cluster. Assuming Space Ranger is installed to /opt/spaceranger-2.1.0
, the process is as follows.
Step 1. Navigate to the Martian runtime's jobmanagers/
directory which contains example job manager templates.
cd /opt/spaceranger-2.1.0/external/martian/jobmanagers tree -L 2 . ├── config.json ├── lsf.template.example ├── pbspro.template.example ├── retry.json ├── sge_queue.py ├── sge.template.example ├── slurm_queue.py ├── slurm.template.example └── torque.template.example
Step 2. Make a copy of the cluster's example Slurm template to slurm.template
in the jobmanagers/
directory.
# command to copy file: cp -v slurm.template.example slurm.template # output looks like this: ‘slurm.template.example’ -> ‘slurm.template’ # command to list files tree -L 2 . ├── config.json ├── lsf.template.example ├── pbspro.template.example ├── retry.json ├── sge_queue.py ├── sge.template.example ├── slurm_queue.py ├── slurm.template ├── slurm.template.example └── torque.template.example
The job submission templates contain several special variables that are substituted by the Martian runtime when each stage is submitted. Specifically, the following variables are expanded when a pipeline is submitting jobs to the cluster.
Variable | Must be present? | Description |
---|---|---|
__MRO_JOB_NAME__ | Yes | Job names composed of the sample ID and stage being executed |
__MRO_THREADS__ | No | Number of threads required by the stage |
__MRO_MEM_GB__ __MRO_MEM_MB__ | No | Amount of memory (in GB or MB) required by the stage |
__MRO_MEM_GB_Per_THREAD__ __MRO_MEM_MB_Per_THREAD__ | No | Amount of memory (in GB or MB) required per thread in multi-threaded stages |
__MRO_STDOUT__ __MRO_STDERR__ | Yes | Paths to the _stdout and _stderr metadata files for the stage |
__MRO_CMD__ | Yes | Bourne shell command to run the stage code |
It is critical that the special variables listed as required are present in the final template you create.
#SBATCH --ntasks=1 --cpus-per-task=__MRO_THREADS__
### Alternatively: --ntasks-per-node=__MRO_THREADS__
Consult with your cluster administrators or help desk to find the combination that works best for single-node, multi-threaded applications on your system.
Considerations for the different job schedulers are listed below.
For SGE cluster, you MUST replace <pe_name>
within the example template to reflect the name of the cluster's multithreaded parallel environment. To view a list of the cluster's parallel environments, use the qconf -spl
command.
The most common modifications to the job submission template include adding additional lines to specify:
- The research group's queue. For example,
#$ -q smith.q
- The account to which jobs will be charged. For example,
#$ -A smith_lab
The most common modifications to the job submission template include adding additional lines to specify:
- Your research group's queue. For example,
#BSUB -q smith_queue
- The account to which your jobs will be charged. For example,
#BSUB -P smith_lab
To run a Space Ranger pipeline in cluster mode, add the --jobmode=slurm
, --jobmode=sge
, or --jobmode=lsf
command line option when using the spaceranger
commands. It is also possible to use --jobmode=<PATH>
, where <PATH>
is the full path to the cluster template file.
To validate that cluster mode is properly configured, follow the same validation instructions given for Space Ranger in the Installation page, but add --jobmode=slurm
, --jobmode=sge
, or --jobmode=lsf
.
spaceranger mkfastq --run=./tiny-bcl --samplesheet=./tiny-sheet.csv --jobmode=slurm ## Expected output Martian Runtime - 4.0.8 Running preflight checks (please wait)... YYYY-MM-DD 12:00:00 [runtime] (ready) ID.HAWT7ADXX.MAKE_FASTQS_CS.MAKE_FASTQS.PREPARE_SAMPLESHEET YYYY-MM-DD 12:00:00 [runtime] (split_complete) ID.HAWT7ADXX.MAKE_FASTQS_CS.MAKE_FASTQS.PREPARE_SAMPLESHEET ...
Once the preflight checks are finished, check the job queue to see the stages queuing up (e.g., using the squeue
command for Slurm).
In the event of a pipeline failure, an error message is displayed.
[error] Pipestance failed. Please see log at: HAWT7ADXX/MAKE_FASTQS_CS/MAKE_FASTQS/MAKE_FASTQS_PREFLIGHT/fork0/chnk0/_errors Saving diagnostics to HAWT7ADXX/HAWT7ADXX.debug.tgz For assistance, upload this file to 10x by running: uploadto10x <your_email> HAWT7ADXX/HAWT7ADXX.debug.tgz
The _errors
file contains a jobcmd error
.
cat AWT7ADXX/MAKE_FASTQS_CS/MAKE_FASTQS/MAKE_FASTQS_PREFLIGHT/fork0/chnk0/_errors jobcmd error: exit status 1
The most likely reason for this failure is an invalid job submission template. This occurs when the job submission via sbatch
, qsub
, or bsub
commands failed.
The "tiny" dataset does not stress the cluster; it may be worthwhile to follow up with a more realistic test using one of our sample datasets.
There are two subtle variants of running spaceranger
pipelines in cluster mode and each has its pitfalls. Check with your cluster administrator to see which approach is compatible with your institution's setup.
- Run
spaceranger
with--jobmode=slurm
on the head node. Cluster mode was originally designed for this use case. However, this approach leavesmrp
andmrjob
running on the head node for the duration of the pipeline, and some clusters impose time limits to prevent long running processes. - Use a job script to submit a
spaceranger
command with--jobmode=slurm
. With this approach,mrp
andmrjob
run on a cluster mode. However, the cluster must allow jobs to be submitted from a compute node to make this viable.
When spaceranger
is run in cluster mode, a single library analysis is partitioned into hundreds and potentially thousands of smaller jobs. The underlying Martian pipeline framework launches each stage job using the sbatch
, qsub
, or bsub
commands when running on the SGE or LSF cluster, respectively. As stage jobs are queued, launched, and completed, the pipeline framework tracks their status using the metadata files that each stage maintains in the pipeline output directory.
Like single server pipelines, cluster-mode pipelines can be restarted after failure. They maintain the same order of execution for dependent stages of the pipeline. All executed stage code is identical to single server mode, and the quantitative results are identical to the limit of each stage's reproducibility.
In addition, the Space Ranger UI can still be used with cluster mode. Because the Martian pipeline framework runs on the node from which the command was issued, the UI will also run from that node.
Stages in the Space Ranger pipelines each request a specific number of cores and memory to aid with resource management. These values are used to prevent oversubscription of the computing system when running pipelines in single server mode. The way CPU and memory requests are handled in cluster mode is defined by the following:
- How the
__MRO_THREADS__
and__MRO_MEM_GB__
variables are used within the job template. - How your specific cluster's job manager schedules resources.
For clusters whose job managers do not support memory requests, it is possible to request memory in the form of cores via the --mempercore
command line option. This option scales up the number of threads requested via the __MRO_THREADS__
variable according to how much memory a stage requires.
For example, given a cluster with nodes that have 16 cores and 128 GB of memory (8 GB per core), the following pipeline invocation command
spaceranger mkfastq --run=./tiny-bcl \ --samplesheet=./tiny-sheet.csv \ --jobmode=slurm \ --mempercore=8
will issue the following resource requests:
- A stage that requires 1 thread and less than 8 GB of memory will pass
__MRO_THREADS__
of1
to the job template. - A stage that requires 1 thread and 12 GB of memory will pass
__MRO_THREADS__
of2
to the job template because (12 GB) / (8 GB/core) = 2 cores. - A stage that requires 2 threads and less than 16 GB of memory will pass
__MRO_THREADS__
of2
to the job template. - A stage that requires 2 threads and 40 GB of memory will pass
__MRO_THREADS__
of5
to the job template because (40 GB) / (8 GB/core) = 5 cores.
As the final bullet point illustrates, this mode can result in wasted CPU cycles and is only provided for clusters that cannot allocate memory as an independent resource.
Every cluster configuration is different. If you are unsure of how your cluster resource management is configured, contact your cluster administrator or help desk.
SGE supports requesting memory via the mem_free
resource natively, although your cluster may have another mechanism for requesting memory. To pass each stage's memory request through to SGE, add a line to your sge.template
that requests mem_free
, h_vmem
, h_rss
, or the custom memory resource defined by your cluster.
cat sge.template ## Expected output #$ -N __MRO_JOB_NAME__ #$ -V #$ -pe threads __MRO_THREADS__ #$ -l mem_free=__MRO_MEM_GB__G #$ -cwd #$ -o __MRO_STDOUT__ #$ -e __MRO_STDERR__ __MRO_CMD__
Note that the h_vmem
(virtual memory) and mem_free/h_rss
(physical memory) represent two different quantities and that Space Ranger stages __MRO_MEM_GB__
requests are expressed as physical memory. Using h_vmem
in your job template may cause certain stages to be unduly killed if their virtual memory consumption is substantially larger than their physical memory consumption. It follows that we do not recommend using h_vmem
.
If you do use h_vmem
in a template, it is recommended that you use the __MRO_MEM_GB_PER_THREAD__
or __MRO_MEM_MB_PER_THREAD__
variables instead of __MRO_MEM_GB__
and __MRO_MEM_MB__
. To determine memory limits for a multicore job, SGE will multiply the number of threads by the value in h_vmem
. The __MRO_MEM_GB__
and __MRO_MEM_MB__
variables already reflect the sum amount of memory across all threads needed to run the job. Using those variables as h_vmem will inflate the memory required for multi-threaded jobs.
LSF supports job memory requests through the -M
and -R [mem=...]
options, but these requests generally must be expressed in MB, not GB. As such, your LSF job template should use the __MRO_MEM_MB__
variable rather than __MRO_MEM_GB__
. For example,
cat bsub.template ## Expected output #BSUB -J __MRO_JOB_NAME__ #BSUB -n __MRO_THREADS__ #BSUB -o __MRO_STDOUT__ #BSUB -e __MRO_STDERR__ #BSUB -R "rusage[mem=__MRO_MEM_MB__]" #BSUB -R span[hosts=1] __MRO_CMD__
Some Space Ranger pipeline stages are divided into hundreds of jobs. By default, the rate at which these jobs are submitted to the cluster is throttled to at most 64 at a time and at least 100 ms between each submission to avoid running into limits on clusters that impose quotas on the total number of pending jobs a user can submit.
If your cluster does not have these limits or is not shared with other users, you can control how the Martian pipeline runner sends job submissions to the cluster by using the --maxjobs
and --jobinterval
parameters.
To increase the cap on the number of concurrent jobs to 200, use the --maxjobs
parameter:
spaceranger count --id=sample ... --jobmode=slurm --maxjobs=200
You can also change the rate limit on how often the Martian pipeline runner sends submissions to the cluster. To add a five-second pause between job submissions, use the --jobinterval
parameter:
spaceranger count --id=sample ... --jobmode=slurm --jobinterval=5000
The job interval parameter is in milliseconds. The minimum allowable value is 1.