Unless otherwise noted:
Log in to the workstation or remote node where cryosparc_master
is installed.
Use the same non-root UNIX user account that runs the cryoSPARC process and was used to install cryoSPARC.
Run all commands on this page in a terminal running bash
Problems with the installation steps are indicated with the some of the following error messages:
"Couldn't connect to host" "Could not resolve host" {"success": false} "tar: This does not look like a tar archive" "Version mismatch! Worker and master versions are not the same. Please update." "An unexpected error has occurred."
Steps
If you have Conda installed, deactivate any active environments.
Check that your LICENSE_ID
environment variable is set correctly with this command
echo $LICENSE_ID
Ensure the output exactly matches the cryoSPARC License ID you were issued over email.
Check your machine's connection to cryoSPARC's license servers at get.cryosparc.com with this curl
command:
curl https://get.cryosparc.com/checklicenseexists/$LICENSE_ID
You should see the message {"success": true}
. If instead you see {"success": false}
, your license is not valid, so please check it has been entered correctly.
If you see an error message like "Couldn't connect to host" or "Could not resolve host" check your Internet connection, firewall or ensure your IT department has the get.cryosparc.com
license server domain whitelisted.
Steps
Reinstall cryoSPARC, following the steps from the "Cannot download cryoSPARC" section
This can happen following a fresh install or recent update.
Steps
In a command line, run cryosparcm status
Check that the output looks like this
----------------------------------------------------------------------------CryoSPARC System master node installed at/home/cryosparcuser/cryosparc/cryosparc_master----------------------------------------------------------------------------cryosparcm process status:app RUNNING pid 13480, uptime 16 days, 3:01:21app_dev STOPPED Not startedcommand_core RUNNING pid 12495, uptime 16 days, 3:01:22command_proxy RUNNING pid 17832, uptime 16 days, 3:01:24command_rtp RUNNING pid 18361, uptime 16 days, 3:01:23command_vis RUNNING pid 19585, uptime 16 days, 3:01:19database RUNNING pid 15879, uptime 16 days, 3:01:30watchdog_dev STOPPED Not startedwebapp RUNNING pid 19805, uptime 16 days, 3:01:18webapp_dev STOPPED Not started----------------------------------------------------------------------------global config variables:export CRYOSPARC_LICENSE_ID="xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx"export CRYOSPARC_MASTER_HOSTNAME="localhost"export CRYOSPARC_DB_PATH="/home/cryosparcuser/cryosparc/cryosparc_database"export CRYOSPARC_BASE_PORT=39000export CRYOSPARC_DEVELOP=falseexport CRYOSPARC_INSECURE=falseexport CRYOSPARC_CLICK_WRAP=true
Check that all items under "cryosparcm process status" that do not end in _dev
are RUNNING
. If any are not, run cryosparcm restart
If any non-_dev
components have a status other than RUNNING
(such as STOPPED
or EXITED
), check their log for errors. For example, this command checks for errors on the database
process:
cryosparcm log database
(Press control C
, then q
to stop logging)
Any error messages here could indicate specific configuration issues and may require re-installing cryoSPARC.
If at any point you see No command 'cryosparcm' found
or command not found: cryosparcm
:
Check that you are on the master node or workstation where cryosparc_master
is installed
Run echo $PATH
and check that it contains <installation directory>/cryosparc_master/bin
$ echo $PATH/home/cryosparcuser/cryosparc/cryosparc_master/bin:/usr/local/cuda-10.1/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
If the web interface is inaccessible, check firewall settings to ensure cryoSPARC's base port number (default 39000
) is exposed for network access
Follow the steps in this section when you see error messages that look like this:
"License is invalid." "License signature invalid." "Could not find local license file. Please re-establish your connection to the license servers." "Local license file is expired. Please re-establish your connection to the license servers." "Token is invalid. Another cryoSPARC instance is running with the same license ID."
Steps
Ensure you entered your license correctly during the installation step.
Check your Internet connection
Check your machine's connection to cryoSPARC's license servers at get.cryosparc.com with this curl
command (substitute <license>
with your unique license ID):
curl https://get.cryosparc.com/checklicenseexists/<license>
You should see the message {"success": true}
If instead you see {"success": false}
, your license is not valid so please check it has been entered correctly.
If you see an error message like "Couldn't connect to host" or "Could not resolve host" check your Internet connection, firewall or ensure your IT department has the get.cryosparc.com
license server domain whitelisted.
If you see a license ID conflict such as this
"Another cryoSPARC instance is running with the same license ID."
Check that no ghost processes are running on all nodes. First stop the running cryoSPARC instance:
cryosparcm stop
For every machine running cryoSPARC (including workers), list all processes associated with cryoSPARC with the following command:
ps -ax | grep cryosparc
If you see any processes that appear associated with cryoSPARC, kill them with the command kill <PID>
, where <PID>
is the process ID number at the beginning of each line.
For example, to kill the first python
process in this process list
PID TTY STAT TIME COMMAND12495 ? Sl 76:06 python -c import cryosparc_command.command_core as serv; serv.start(port=39002)15879 ? Sl 379:49 mongod --dbpath /home/cryosparcuser/cryosparc/cryosparc_database --port 39001 --replSet meteor17832 ? S 0:02 python -c import cryosparc_command.command_proxy as serv; serv.start(port=39004)18361 ? Sl 288:40 python -c import cryosparc_command.command_rtp as serv; serv.start(port=39005)19585 ? Sl 46:34 python -c import cryosparc_command.command_vis as serv; serv.start(port=39003)
Run
kill 12495
Alternatively, if you have the killall
program installed, kill all running python processes with one of these commands:
killall -u $(whoami) python# orsudo killall -u <cryosparcuser> python
If the current Linux user runs cryoSPARC, use the first command. Otherwise use the second command, substituting <cryosparcuser>
with the username that runs cryoSPARC.
Then start cryoSPARC on the master note or workstation:
cryosparcm start
Follow these steps when the cryoSPARC web interface is up and running normally and jobs may be created but do not run. These error messages may indicate this issue:
"list index out of range" "Could not resolve hostname ... Name or service not known"
A job that never changes from Queued
or Launched
status may also indicate this.
Steps
Ensure at least one worker is connected to the master. See the Installation page for details.
Visit the /resources
page to see what lanes are available
Check that all non-development cryoSPARC processes are running with the cryosparcm status
command
(For master/worker setups) check that SSH is configured between the master and worker machines.
Check the log for the command_core
log to find any application error messages
cryosparcm log command_core
(press Control + C
on the keyboard to exit when finished)
If applicable, check that the cluster submission script is correct
Refresh job types and reload cryoSPARC:
cryosparcm cli "refresh_job_types()"cryosparcm cli "reload()"
On the workstation or master node run:
cryosparcm forcedeps
On worker nodes run:
cryosparcw forcedeps
Restart cryoSPARC:
cryosparcm restart
Then Clear the job and re-run it.
When a job fails, its job card in the interface turns red and the bottom of the job log includes an error message with the textTraceback (most recent call last)
Common failure reasons include:
Invalid or unspecified input slots
Invalid or unspecified required parameters, including file/folder paths
Incorrectly set up GPU (e.g., running a job on a node without enough GPUs or CUDA drivers not installed)
Another process taking up memory on a GPU
Cache not set up correctly for a worker
Common job failure error messages:
"AssertionError: Child process with PID ... has terminated unexpectedly!"
Common error messages that indicate incorrectly configured GPU drivers:
"cuInit failed: unknown error" "no CUDA-capable device is detected" "cuMemHostAlloc failed: OS call failed or operation not supported on this OS" "cuCtxCreate failed: invalid device ordinal kernel.cu ... error: identifier "__shfl_down_sync" is undefined
Common error messages that indicate not enough GPU memory:
"cuMemAlloc failed: out of memory" "cuArrayCreate failed: out of memory" "cufftAllocFailed"
Steps
Ensure a CUDA version ≥10 is installed and running on the workstation or each worker node
Check the GPU configuration on the workstation or node where the job runs on. Log into that machine and navigate to the cryoSPARC installation directory. Run the cryosparcw gpulist
command:
cd /path/to/cryosparc_workerbin/cryosparcw gpulist
Run nvidia-smi
to check the GPU and CUDA driver status.
Check that CUDA version is within the allowed range. cryoSPARC supports CUDA versions ≥10
Check that no other processes are using GPU memory. cryoSPARC-related process appear with process name "python"
If you don't recognize the processes using GPU memory, run kill <PID>
, substituting <PID>
with the value under the Processes PID column
Check the Job log: Select the Job card in the cryoSPARC interface and press the Spacebar on your keyboard to see the log. Scroll down to the bottom and look for the failure reasons in red
Clear the job and select the "Building" status badge on the job card to enter the Job Builder
If the job failed with a GPU-related error and multiple GPUs are available, try running the job on a different GPU. Press Queue, switch to the "Run on specific GPU" Queue type and select one or more GPUs
If the job failed with AssertionError: Non-optional inputs from the following input groups and their slots are not connected
then expand any input groups connected to the job. Missing required slots appear in red
Check the job parameters To learn about setting specific parameters, hover over or touch the parameter names in the Job Builder to see a description of what they do
Find the target job type in this guide's Job Reference for more detailed descriptions of expected input slots and parameters.
Reduce the box-size of extracted particles. Some jobs need to fit thousands of particles in GPU memory at a time, and larger box sizes exceed GPU memory limits. Either extract with a smaller box size or with the Downsample Particles job
Refresh job types and reload cryoSPARC:
cryosparcm cli "refresh_job_types()"cryosparcm cli "reload()"
Look for extended error information with the cryosparcm joblog
command (press Control + C
on the keyboard to exit when finished)
On occasion, a job fails due to an error in the cryoSPARC code (bug). When a bug is discovered, the cryoSPARC team promptly patches the issue and makes it available in the next update.
If you find a new bug, see the Additional Help section for advice.
By nature, the size of cryo-EM datasets can take a long time to process with sub-optimal hardware or parameters. Here are some facilities that cryoSPARC provides for increasing speed/performance.
Connect workers with SSD cache enabled. This speeds up processing for extracted particles during 2D Classification, ab-initio reconstruction, refinement and more. Ensure the "Cache particle images on SSD" parameter is enabled under "Compute settings" for the target particle-processing job
Some jobs (motion correction, ctf estimation, 2D classification) can run on multiple GPUs. If your hardware supports it, increase the number of GPUs to parallelize over under
Extracted particles with large box sizes (relative to their pixel size) take a long time to process. Consider Fourier-cropping (or "binning") the extracted particle blobs with the Downsample Particles job
Minimize the number of processes using system resources on the workstation or worker nodes
Check for zombie processes on worker machines. The process is similar to the steps under "Another cryoSPARC instance is running with the same license ID" under the License error or license not found section
Cancel the job, clear and re-queue
File "/u/cryosparc/cryosparc_worker/cryosparc_compute/jobs/runcommon.py", line 1711, in run_with_except_hookrun_old(*args, **kw)File "cryosparc_worker/cryosparc_compute/engine/cuda_core.py", line 129, in cryosparc_compute.engine.cuda_core.GPUThread.runFile "cryosparc_worker/cryosparc_compute/engine/cuda_core.py", line 130, in cryosparc_compute.engine.cuda_core.GPUThread.runFile "cryosparc_worker/cryosparc_compute/engine/engine.py", line 997, in cryosparc_compute.engine.engine.process.workFile "cryosparc_worker/cryosparc_compute/engine/engine.py", line 106, in cryosparc_compute.engine.engine.EngineThread.load_image_data_gpuFile "cryosparc_worker/cryosparc_compute/engine/gfourier.py", line 33, in cryosparc_compute.engine.gfourier.fft2_on_gpu_inplaceFile "/u/cryosparc/cryosparc_worker/deps/anaconda/envs/cryosparc_worker_env/lib/python3.7/site-packages/skcuda/fft.py", line 102, in __init__capability = misc.get_compute_capability(misc.get_current_device())File "/u/cryosparc/cryosparc_worker/deps/anaconda/envs/cryosparc_worker_env/lib/python3.7/site-packages/skcuda/misc.py", line 254, in get_current_devicereturn drv.Device(cuda.cudaGetDevice())File "/u/cryosparc/cryosparc_worker/deps/anaconda/envs/cryosparc_worker_env/lib/python3.7/site-packages/skcuda/cudart.py", line 767, in cudaGetDevicecudaCheckStatus(status)File "/u/cryosparc/cryosparc_worker/deps/anaconda/envs/cryosparc_worker_env/lib/python3.7/site-packages/skcuda/cudart.py", line 565, in cudaCheckStatusraise eskcuda.cudart.cudaErrorInsufficientDriver
This error usually indicates the NVIDIA driver you're using isn't compatible with the GPU and CUDA Toolkit version installed. This issue can be fixed by installing the latest NVIDIA Driver from the driver download page.
Related Discussion Forum post where a user encountered this error.
Once installed, you can recompile pyCUDA by using the newcuda
command: cryosparcw reference.
You can get additional help through the cryoSPARC Discussion Forum where users post questions, tips and bug reports not dealt with above, or by contacting us directly.
Search for keywords related to the specific issue on the cryoSPARC Discussion Forum. If no related discussions exist, please create a new post.
If assistance is still required, get in touch with the cryoSPARC team directly via the Questions and Support page.