This file keeps track of all notable changes to the Cluster Agent.
- Added support to set slurm restd version dynamically
- Added a timeout to the LDAP connection
- Added the ability to recover the agent when a connection to the LDAP server fails
- Capture rejection message from Slurm
- Capture failure message from Slurm jobs
- Removed job-script parser to extract the name and values of the SBATCH parameters contained in the file.
- Added logic to pull execution_parameters from job submissions and use them as job properties when jobs are submitted.
- Fix SBATCH parameter --exclusive
- Fix logic that checks if both SLURMRESTD_JWT_KEY_PATH and SLURMRESTD_JWT_KEY_STRING parameters are set
- Introduce support to generate the Slurmrestd JWT by Python code instead of calling
scontrol token
.
- Fix report messages for rejected job submissions.
- Add supporting file functionality to job submissions.
- Update code to fit Keycloak OIDC authentication instead of Auth0.
- Update patch version for publishing on Pypi.
- Added job-script parser to extract the name and values of the SBATCH parameters contained in the file.
- Added a two-way mapping to translate parameters' names between the SBATCH namespace and the Slurm Rest API namespace.
- Added execution directory support.
- Add environment option for sentry.
- Disabled cluster data collectors.
- Moved LDAP support into
slurm_user
module. - Made slurm user mapping more extensible.
- Fixed username mapping for NTLM auth type.
- Added jobbergate section for retrieving, submitting, and updating jobs from Jobbergate
- Added support for looking up usernames from LDAP via email from auth token
- Added unit tests for jobbergate section
- removed dependency on FastAPI and Uvicorn;
- removed autoscheduling from the agent;
- implemented a function to call sequentially the functions that collect slurmrestd data.
- implemented logic to authenticate against the Cluster API by Auth0 tokens;
- changed logic to issue Slurmrestd JWT;
- removed request module;
- created identity module to handle client logics for the APIs (Cluster APi and Slurmrestd).
- bump project name from armada-agent to cluster-agent.
- fixed backend's partition route.
- changed the logging stack to use Loguru.
- changed function to issue slurm JWT token without passing the username explicitly.
- implemented the core functionalities of the agent: scrape slurmrestd for diagnostics, jobs, partitions and nodes data;
- implemented GitHub workflows for testing and linting the code, as well as publishing it to AWS CodeArtifact.