Ejemplo n.º 1
0
def test_get_available_nodes_central_mode_remove_libE_proc():
    mynode = socket.gethostname()
    nodelist_in = [
        'knl-0020', 'knl-0021', 'knl-0022', 'knl-0036', 'knl-0137', 'knl-0138',
        'knl-0139', 'knl-1234'
    ]
    with open('worker_list', 'w') as f:
        for i, node in enumerate(nodelist_in):
            f.write(node + '\n')
            if i == 3:
                f.write(mynode + '\n')

    resources = Resources(central_mode=True)

    #Now mock up some more stuff - so consistent

    #Spoof current process as each worker and check nodelist.
    resources.num_workers = 8
    exp_out = [['knl-0020'], ['knl-0021'], ['knl-0022'], ['knl-0036'],
               ['knl-0137'], ['knl-0138'], ['knl-0139'], ['knl-1234']]
    for wrk in range(resources.num_workers):
        resources.workerID = wrk + 1
        local_nodelist = resources.get_available_nodes()
        assert local_nodelist == exp_out[
            wrk], "local_nodelist returned does not match expected"

    #Spoof current process as each worker and check nodelist.
    resources.num_workers = 4
    exp_out = [['knl-0020', 'knl-0021'], ['knl-0022', 'knl-0036'],
               ['knl-0137', 'knl-0138'], ['knl-0139', 'knl-1234']]
    for wrk in range(resources.num_workers):
        resources.workerID = wrk + 1
        local_nodelist = resources.get_available_nodes()
        assert local_nodelist == exp_out[
            wrk], "local_nodelist returned does not match expected"

    #Spoof current process as each worker and check nodelist.
    resources.num_workers = 1
    exp_out = [[
        'knl-0020', 'knl-0021', 'knl-0022', 'knl-0036', 'knl-0137', 'knl-0138',
        'knl-0139', 'knl-1234'
    ]]
    for wrk in range(resources.num_workers):
        resources.workerID = wrk + 1
        local_nodelist = resources.get_available_nodes()
        assert local_nodelist == exp_out[
            wrk], "local_nodelist returned does not match expected"

    #Test the best_split algorithm
    resources.num_workers = 3
    exp_out = [['knl-0020', 'knl-0021', 'knl-0022'],
               ['knl-0036', 'knl-0137', 'knl-0138'], ['knl-0139', 'knl-1234']]
    for wrk in range(resources.num_workers):
        resources.workerID = wrk + 1
        local_nodelist = resources.get_available_nodes()
        assert local_nodelist == exp_out[
            wrk], "local_nodelist returned does not match expected"

    os.remove('worker_list')
Ejemplo n.º 2
0
def test_get_available_nodes_central_mode():
    os.environ[
        "LIBE_RESOURCES_TEST_NODE_LIST"] = "knl-[0020-0022,0036,0137-0139,1234]"
    resources = Resources(nodelist_env_slurm="LIBE_RESOURCES_TEST_NODE_LIST",
                          central_mode=True)

    #Now mock up some more stuff - so consistent

    #Spoof current process as each worker and check nodelist.
    resources.num_workers = 8
    exp_out = [['knl-0020'], ['knl-0021'], ['knl-0022'], ['knl-0036'],
               ['knl-0137'], ['knl-0138'], ['knl-0139'], ['knl-1234']]
    for wrk in range(resources.num_workers):
        resources.workerID = wrk + 1
        local_nodelist = resources.get_available_nodes()
        assert local_nodelist == exp_out[
            wrk], "local_nodelist returned does not match expected"

    #Spoof current process as each worker and check nodelist.
    resources.num_workers = 4
    exp_out = [['knl-0020', 'knl-0021'], ['knl-0022', 'knl-0036'],
               ['knl-0137', 'knl-0138'], ['knl-0139', 'knl-1234']]
    for wrk in range(resources.num_workers):
        resources.workerID = wrk + 1
        local_nodelist = resources.get_available_nodes()
        assert local_nodelist == exp_out[
            wrk], "local_nodelist returned does not match expected"

    #Spoof current process as each worker and check nodelist.
    resources.num_workers = 1
    exp_out = [[
        'knl-0020', 'knl-0021', 'knl-0022', 'knl-0036', 'knl-0137', 'knl-0138',
        'knl-0139', 'knl-1234'
    ]]
    for wrk in range(resources.num_workers):
        resources.workerID = wrk + 1
        local_nodelist = resources.get_available_nodes()
        assert local_nodelist == exp_out[
            wrk], "local_nodelist returned does not match expected"

    #Test the best_split algorithm
    resources.num_workers = 3
    exp_out = [['knl-0020', 'knl-0021', 'knl-0022'],
               ['knl-0036', 'knl-0137', 'knl-0138'], ['knl-0139', 'knl-1234']]
    for wrk in range(resources.num_workers):
        resources.workerID = wrk + 1
        local_nodelist = resources.get_available_nodes()
        assert local_nodelist == exp_out[
            wrk], "local_nodelist returned does not match expected"
Ejemplo n.º 3
0
def test_get_available_nodes_distrib_mode_host_not_in_list():
    os.environ[
        "LIBE_RESOURCES_TEST_NODE_LIST"] = "knl-[0020-0022,0036,0137-0139,1234]"
    resources = Resources(nodelist_env_slurm="LIBE_RESOURCES_TEST_NODE_LIST",
                          central_mode=False)

    #Spoof current process as each worker and check nodelist.
    resources.num_workers = 8
    exp_out = [['knl-0020'], ['knl-0021'], ['knl-0022'], ['knl-0036'],
               ['knl-0137'], ['knl-0138'], ['knl-0139'], ['knl-1234']]

    # Test running distributed mode without current host in list.
    resources.workerID = 2
    try:
        local_nodelist = resources.get_available_nodes()
    except:
        assert 1
    else:
        assert 0
Ejemplo n.º 4
0
def test_get_available_nodes_distrib_mode_uneven_split():
    mynode = socket.gethostname()
    nodelist_in = [
        'knl-0020', 'knl-0021', 'knl-0022', 'knl-0036', 'knl-0137', 'knl-0138',
        'knl-0139', 'knl-1234'
    ]
    with open('worker_list', 'w') as f:
        for i, node in enumerate(nodelist_in):
            f.write(node + '\n')
            if i == 4:
                f.write(mynode + '\n')

    resources = Resources(central_mode=False)
    resources.num_workers = 2

    # May not be at head of list - should perhaps be warning or enforced
    resources.workerID = 2
    exp_out = ['knl-0137', mynode, 'knl-0138', 'knl-0139']
    local_nodelist = resources.get_available_nodes()
    assert local_nodelist == exp_out, "local_nodelist returned does not match expected"
    os.remove('worker_list')
Ejemplo n.º 5
0
def test_get_available_nodes_distrib_mode():
    mynode = socket.gethostname()
    #nodelist_in = ['knl-0020', 'knl-0021', 'knl-0022', 'knl-0036', 'knl-0137','knl-0138', 'knl-0139', 'knl-1234']
    nodelist_in = [
        'knl-0020', 'knl-0021', 'knl-0022', 'knl-0036', 'knl-0137', 'knl-0138',
        'knl-0139'
    ]
    with open('worker_list', 'w') as f:
        for i, node in enumerate(nodelist_in):
            f.write(node + '\n')
            if i == 3:
                f.write(mynode + '\n')

    resources = Resources(central_mode=False)

    #Spoof current process as each worker and check nodelist.
    resources.num_workers = 8

    #Test workerID not in local_nodelist
    resources.workerID = 4
    try:
        local_nodelist = resources.get_available_nodes()
    except:
        assert 1
    else:
        assert 0

    resources.workerID = 5
    exp_out = [mynode]
    local_nodelist = resources.get_available_nodes()
    assert local_nodelist == exp_out, "local_nodelist returned does not match expected"

    resources.num_workers = 1
    resources.workerID = 1
    exp_out = [
        'knl-0020', 'knl-0021', 'knl-0022', 'knl-0036', mynode, 'knl-0137',
        'knl-0138', 'knl-0139'
    ]
    local_nodelist = resources.get_available_nodes()
    assert local_nodelist == exp_out, "local_nodelist returned does not match expected"

    resources.num_workers = 4
    resources.workerID = 3
    exp_out = [mynode, 'knl-0137']
    local_nodelist = resources.get_available_nodes()
    assert local_nodelist == exp_out, "local_nodelist returned does not match expected"

    #Sub-node workers
    resources.num_workers = 16

    resources.workerID = 9
    exp_out = [mynode]
    local_nodelist = resources.get_available_nodes()
    assert local_nodelist == exp_out, "local_nodelist returned does not match expected"

    resources.workerID = 10
    exp_out = [mynode]
    #import pdb; pdb.set_trace()
    local_nodelist = resources.get_available_nodes()
    assert local_nodelist == exp_out, "local_nodelist returned does not match expected"
    os.remove('worker_list')