slave_health_for_type.py

import os
from datetime import datetime

import psycopg2

conn = psycopg2.connect(os.environ['PGCONNECTIONSTRING'])
cur = conn.cursor()

# this should be a list that TC could be using that is not currently allocated
# to buildbot
machines = ["t-yosemite-r7-{:04d}".format(i) for i in range(300)]
machine_data = {}

def get_machine_state(worker_id):
    machine_data = {
        'elapsed_since_job': None,
        'elapsed_since_job_secs': 0,
        'elapsed_on_job': 0,
        'slave_state': 'no jobs',
        'starttime': None, # not sure what this is
        'slave_class': 'test',
        'row_class': 'no jobs',
    }

    # we only care about the last 2 tasks that a worker might have executed
    cur.execute("select state,scheduled,started,resolved from tasks where worker_id = %s order by started desc limit 1", (machine,))

    if cur.rowcount == 0:
        return machine_data

    for r in cur:
        if r[0] == 'running':
            machine_data['row_class'] = 'working'
            machine_data['slave_state'] = 'working'
            machine_data['elapsed_on_job'] = datetime.now() - started
            return machine_data

        # if the task is not running, assumed exception or resolved
        since = (datetime.utcnow() - r[3]).total_seconds()
        m, s = divmod(since, 60)
        h, m = divmod(m, 60)
        machine_data['elapsed_since_job'] = "{%d:%02d:%02d}" % (h, m, s)
        machine_data['elapsed_since_job_secs'] = since
        if since/3600.0 > 5:
            machine_data['row_class'] = machine_data['slave_state'] = 'broken'

        return machine_data

try:
    for machine in machines:
        machine_data[machine] = get_machine_state(machine)
    print machine_data
finally:
    cur.close()
    conn.close()