Welcome to ceibacli’s documentation!¶
ceiba-cli¶
command line interface to interact with the insilico web server. See the documentation and blog post.
Installation¶
To install ceiba-cli, do:
pip install git+https://github.com/nlesc-nano/ceiba-cli.git@main
Contributing¶
If you want to contribute to the development of ceiba-cli, have a look at the contribution guidelines.
License¶
Copyright (c) 2020-2021, Netherlands eScience Center
Licensed under the Apache License, Version 2.0 (the “License”); you may not use this file except in compliance with the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an “AS IS” BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.
Credits¶
This package was created with Cookiecutter and the NLeSC/python-template.
Authentication¶
Generate an OAuth token¶
You need to generate an OAuth token from GitHub in order to login into the application! For doing so, you should:
Go to github tokens and click the
Generate new tokenbutton.Provide your GitHub password when prompted
Fill in a description for the token, for example, ceiba access token.
Do not enable any scope therefore the token will grant read-only access to the app
Click
Generateat the bottom. Make sure to copy its value because you will need it to login!
Usage¶
The ceibacli command line interface offers four actions to interact with the Ceiba web service. You can check them by trying the following command in your terminal:
user> ceibacli --help
You should see something similar to:
usage: ceibacli [-h] [--version] {login,add,compute,report,query,manage} ...
positional arguments:
{login,add,compute,report,query,manage}
Interact with the properties web service
login Log in to the Insilico web service
add Add new jobs to the database
compute Compute available jobs
report Report the results back to the server
query Query some properties from the database
manage Change jobs status
optional arguments:
-h, --help show this help message and exit
--version show program's version number and exit
After running one of the previous commands a log file named ceibacli_output.log
is generated.
Login¶
We certainly want to restrict who can access and modify the data. Therefore users are required to login with the web service. For doing so, you should have a GitHub account, then you need to request a read-only token from GitHub personal access token service.
Once you have a read-only GitHub token, you can login into the web service like:
ceibacli login -w http://YourCeibaInstance:8080/graphql -t Your_token
How does it work?¶
The Ceiba server will contact GitHub and will check if you are a known user there.
Add (for Administrators)¶
The add command is an adminstrative action to add new jobs into the database.
To add jobs you need to run the following command in the terminal:
ceibacli add -w http://yourCeibaInstance:8080/grapqhl -c collection_name -j Path/to/jobs.json
Where the -w option is the web service URL. the collection_name is the collection where the data is going to be stored. Finally, the -j options is the path to the JSON file containing the jobs as an array of JSON objects. See the next Jobs File section for further information.
Jobs File¶
The job file is a list of json objects, like:
[
{
"type": "awesome_simulation_1",
"parameters": {
"value": 3.14
}
},
{
"type": "awesome_simulation_2",
"parameters": {
"value": 2.187
}
}
]
Each job is a JSON object with the parameters to perform the simulation.
How does it work?¶
The add command will read each job in the JSON jobs file. For each job it will generate a unique identifiers. Then, the jobs and their identifier will be stored a collection named job_your_collection_name.
Compute¶
The compute command ask the web service for available jobs that need to be run.
To run some jobs you need to type in the terminal:
ceibacli compute -i input_compute.yml
Where the input_compute.yml is an file in YAML format containing the Compute Input File metadata.
The compute command takes the user’s input, request some available job and Job Scheduling those jobs using the information provided by the user.
Compute Input File¶
The input file contains the following mandatory keywords:
# Web service URL
web: "http://YourCeibaInstance:8080/graphql"
# Name of the collection to compute
collection_name: "simulation_name"
# Command use to run the workflow
command: compute_properties
Other optional keywords are:
# Configuration of the job scheduler
scheduler:
"none"
# Path to the directory where the calculations are going to run (default: workdir_ceibacli)
workdir:
/path/to/workdir
# Number of jobs to request and run (default: 10)
max_jobs:
5
Job Scheduling¶
Most of the scientific simulation are usually perform in supercomputers that use a
job scheduler. ceiba-cli supports two of the most popular ones: SLURM.
If you choose a scheduler different from none, ceiba-cli will automatically contact
the job scheduler with the options that you have provided. Below you can find a description
of the available options:
# Job scheduler. Of of "none" or "slurm" (default: none)
scheduler:
slurm
# Number of computing nodes to request (default: 1)
nodes:
1
# Number of CPUs per task (default: None)
cpus_per_task:
48
# Total time to request ind "days:hours:minutes" format (default: 1day)
walltime:
"01:00:00"
# Partion name (queue's name) where the job is going to run (default: None)
partion_name:
"short"
You can alternatively provide a string with all the options for the queue system like,
scheduler:
slurm
# String with user's Configuration
free_format: "#!/bin/bash
#SBATCH -N 1
#SBATCH -t 00:15:00
....
"
Job State¶
The user’s requested jobs are initially marked as RESERVERED, in the web service to
avoid conflicts with other users. Then, if the jobs are sucessfully scheduled they
are marked as RUNNING. If there is a problem during the scheduling or subsequent
running step the job would be marked as FAILED.
Report¶
The report command send the results of the jobs computed by the user to
the web service. You can also send data that is not associated to any job to the server.
In the last case, the results don’t have all the metadata associated with a job in the server,
for example because it has been previously computed or computed in another facility.
To report the results you need to type in the terminal:
ceibacli report -w http://yourCeibaInstance:8080/grapqhl
Or if you want to have more control over what is reported you can provide an input file like:
ceibacli report -i input_report.yml
Where the input_compute.yml is an file in YAML format containing the Report results from a job metadata.
You can also report results without associated jobs, follow the Report results without associated jobs.
Report results from a job¶
If the results that you want to report where computed with the ceibacli compute command, you can optionally provide the following input:
# Path to the Folder where the jobs run (default "workdir_ceibacli")
path_results: "workdir_ceibacli"
# Pattern to search for the result files (default "results*csv")
output: "results*csv"
# Pattern to search for the input files (default "inputs*json")
input: "inputs*json"
# If the data is already in server you can either:
# KEEP the old data
# OVERWRITE and discard the old data
# MERGE the new and the old data (Default)
# APPEND new data at the end of the old data array
# Default = KEEP
duplication_policy: "KEEP"
Check the Large objects data storage for further information on saving large output files.
Report results without associated jobs¶
Sometimes you have some results that you have previously computed and you want to share them with your colleagues. You can upload those results into the database very similarly to the previous section, but you need to provide an additional keyword:
has_metadata: False
You also need to provide the path_results and the output to look for. The has_metadata
indicates to Ceiba-cli that the results that you want to report don’t have metadata about how the
results where computed.
How does it work?¶
The library enters the path_results and search recursively all the files and
directories name like job_*. In each subfolder, apart from the
computed data (specificied with the pattern keyword), the report command
would try to collect the metadata associated with the job in a files named
metadata.yml containing the following information:
job_id: 1271269411
property:
collection_name: awesome_data
id: 76950
Without the metadata no data is reported back to the server.
Reporting data without associated jobs¶
Large objects data storage¶
For many simulation it is desirable to store the output plain data and/or the binary checkpoints. Those files can be used to retrieve data that is not available in the database or to restart a calculation to perform further computations.
Those large objects are not suitable for storage in a database but fortunately there are technologies like swift openstack that allows to store these kind of data in an efficient and safely way.
In order to storage large output you need to provide in the yaml file the following keywords:
large_objects:
# URL to the datastorage service
web: "http://large_scientific_data_storage.pi"
# The large file(s) to search for
patterns: ["output*hdf5"]
Note
Installing, deploying an mantaining a swift openstack data storage service is a nontrivial task. Therefore it is recommended to request access to this service to a provider. Be aware that IT COSTS MONEY to maintain the service running in a server!
The large files and their corresponding metadata are going to be stored in the swift collection. using the same
collection_namethat has been specified in the How does it work?.
Query¶
The query actions requests some data from the web service
and writes the requested data in a csv file.
- There are currently two possible query actions:
request what collections are available
request a single collection
To request what collections are available you just need to run the following command:
ceibacli query -w http://yourCeibaInstance:8080/grapqhl
Previous command will ouput something similar to:
Available collections:
name size
simulation1 3
simulation2 42
....
In the previous name indicates the actual collections’ names and size how many datasets are stored
in that particular collection.
To request all the datasets available in a given collection, you just need to run the following command:
ceibacli query -w http://yourCeibaInstance:8080/grapqhl -c simulation2
That command will write into your current work directory a file called simulation2.csv
containing the properties in the requested collection.
Manage (For administrators)¶
The manage command is an adminstrative action to change the jobs status. For example,
jobs that have been marked as RESERVED or RUNNING for a long period of time
can be marked again as AVAILABLE if the user doesn’t report the results.
To change the jobs status you need to type in the terminal:
ceibacli manage -i input_manage.yml
Where the input_manage.yml is an file in YAML format containing the Manage Input File specification.
Manage Input File¶
The following snippet represent an input example for the manage action:
# Web service URL
web: "http://YourCeibaInstance:8080/graphql"
# Target collection to change job status
collection_name: "example_collection"
# Metadata to change jobs status
change_status:
old_status: RUNNING
new_status: AVAILABLE
expiration_time: 24 # one day
How does it work?¶
ceiba-cli will research in the collection_name for all the jobs with old_status then
it will check if those jobs have been scheduled before the expiration_time. If
the jobs have expired, ceiba-cli will marked the expired jobs with the new_status.