ores.scheduler.job_definition

A job the scheduler fires on a cron expression. The row carries the schedule (schedule_expression), what to run when it fires, and whether it is active. action_type selects the behaviour: execute_sql runs the SQL in command, and nats_publish publishes the subject and body carried in action_payload.

The table is bi-temporal and audited (see projects/ores.sql/create/scheduler/scheduler_job_definitions_create.sql): it carries version, the four audit columns and the valid_from=/=valid_to pair with the GIST exclusion and the delete rule, so the model takes the ordinary audited shape and needs no shape flag.

tenant_id is nullable. A job may belong to no tenant, because the scheduler fires system jobs from a NULL-tenant row: the MQ statistics scrape and the compute stale-result reaper are both such rows. The model binds uuid-identified-lookup for its UUID surrogate key, its tenant scope and its standard presentation tier, and states nullable_tenant_id itself, which that profile leaves to the model.

job_name is the natural key and is unique within its tenant; id is the surrogate. Uniqueness is what the component's upsert path relies on: a job that arrives under an existing name updates that row in place instead of adding a second one.

The component's operational views — the global job-instance list and the live scheduler status — are operations rather than entity verbs, and are modelled in ores.scheduler.scheduling_operations.

#include <chrono>
#include <optional>
#include <string>
#include <boost/uuid/uuid.hpp>

/**
 * @brief Reads every active job definition, across all tenants.
 *
 * The scheduler loop fires system jobs as well as tenant ones, so it
 * cannot read through the tenant-scoped CRUD set.
 */
std::vector<domain::job_definition> read_all_active(context ctx);

std::vector<domain::job_definition>
job_definition_repository::read_all_active(context ctx) {
    BOOST_LOG_SEV(lg(), debug) << "Reading all active job definitions.";
    static const auto max(make_timestamp(MAX_TIMESTAMP, lg()));
    const auto query = sqlgen::read<std::vector<job_definition_entity>> |
                       where("is_active"_c == true && "valid_to"_c == max.value());

    return execute_read_query<job_definition_entity, domain::job_definition>(
        ctx,
        query,
        [](const auto& entities) { return job_definition_mapper::map(entities); },
        lg(),
        "Reading all active job definitions.");
}