condorstartd

condor_startd is the HTCondor daemon that runs on every execute host in an HTCondor pool. It is responsible for managing local resources and launching and supervising user jobs that have been matched to the host. The startd communicates with the central management components, advertising the host’s capabilities to the collector and participating in the resource negotiation process via the negotiator and schedd.

When a job is selected by the schedd, the startd is tasked with starting the job on the execute host. It delegates the actual launch to a helper process called condor_starter, which creates the execution sandbox, initializes the environment, handles input/output, and applies resource limits. The startd then monitors the running job and can suspend, resume, preempt, or terminate it according to the configured policies.

The startd supports multiple execution slots on a single machine and enforces limits on CPU, memory, and disk usage. It provides accounting information to the collector and integrates with HTCondor’s checkpointing and migration features where available. The daemon also coordinates with the master to ensure reliable startup, monitoring, and cleanup of jobs and resources.

Configuration and operation: Administrators configure the startd through HTCondor configuration files, enabling or tuning advertised resources and policy behavior. The startd runs under the HTCondor master service and can be managed with standard system tools and commands.

See also: HTCondor, condor_schedd, condor_collector, condor_negotiator, condor_master.