def nuke(ctx, log, should_unlock, sync_clocks=True, reboot_all=True,
noipmi=False):
from teuthology.parallel import parallel
from teuthology.lock import list_locks
total_unnuked = {}
targets = dict(ctx.config['targets'])
if ctx.name:
log.info('Checking targets against current locks')
locks = list_locks(ctx)
#Remove targets who's description doesn't match archive name.
for lock in locks:
for target in targets:
if target == lock['name']:
if ctx.name not in lock['description']:
del ctx.config['targets'][lock['name']]
log.info('Not nuking %s because description doesn\'t match', lock['name'])
with parallel() as p:
for target, hostkey in ctx.config['targets'].iteritems():
p.spawn(
nuke_one,
ctx,
{target: hostkey},
log,
should_unlock,
sync_clocks,
reboot_all,
ctx.config.get('check-locks', True),
noipmi,
)
for unnuked in p:
if unnuked:
total_unnuked.update(unnuked)
if total_unnuked:
log.error('Could not nuke the following targets:\n' + '\n '.join(['targets:', ] + yaml.safe_dump(total_unnuked, default_flow_style=False).splitlines()))
def get_arch(config):
for yamlfile in config:
y = yaml.safe_load(file(yamlfile))
machine_type = y.get('machine_type')
if machine_type:
locks = lock.list_locks()
for machine in locks:
if machine['type'] == machine_type:
arch = machine['arch']
return arch
return None
def get_arch(config):
for yamlfile in config:
y = yaml.safe_load(file(yamlfile))
machine_type = y.get('machine_type')
if machine_type:
locks = lock.list_locks()
for machine in locks:
if machine['type'] == machine_type:
arch = machine['arch']
return arch
return None
def get_arch(machine_type):
"""
Based on a given machine_type, return its architecture by querying the lock
server. Sound expensive? It is!
:returns: A string or None
"""
locks = lock.list_locks()
for machine in locks:
if machine['type'] == machine_type:
arch = machine['arch']
return arch
return None
def lock_machines(ctx, config):
log.info('Locking machines...')
assert isinstance(config, int), 'config must be an integer'
while True:
# make sure there are enough machines up
machines = lock.list_locks(ctx)
if machines is None:
if ctx.block:
log.warn('error listing machines, trying again')
time.sleep(20)
continue
else:
assert 0, 'error listing machines'
num_up = len(filter(lambda machine: machine['up'], machines))
assert num_up >= config, 'not enough machines are up'
# make sure there are machines for non-automated jobs to run
num_free = len(
filter(lambda machine: machine['up'] and machine['locked'] == 0,
machines))
if num_free < 6 and ctx.owner.startswith('scheduled'):
if ctx.block:
log.info('waiting for more machines to be free...')
time.sleep(10)
continue
else:
assert 0, 'not enough machines free'
newly_locked = lock.lock_many(ctx, config, ctx.owner)
if len(newly_locked) == config:
ctx.config['targets'] = newly_locked
log.info('\n '.join([
'Locked targets:',
] + yaml.safe_dump(ctx.config['targets'],
default_flow_style=False).splitlines()))
break
elif not ctx.block:
assert 0, 'not enough machines are available'
log.warn('Could not lock enough machines, waiting...')
time.sleep(10)
try:
yield
finally:
if ctx.summary.get('success', False):
log.info('Unlocking machines...')
for machine in ctx.config['targets'].iterkeys():
lock.unlock(ctx, machine, ctx.owner)
def lock_machines(ctx, config):
log.info('Locking machines...')
assert isinstance(config, int), 'config must be an integer'
while True:
# make sure there are enough machines up
machines = lock.list_locks(ctx)
if machines is None:
if ctx.block:
log.warn('error listing machines, trying again')
time.sleep(20)
continue
else:
assert 0, 'error listing machines'
num_up = len(filter(lambda machine: machine['up'], machines))
assert num_up >= config, 'not enough machines are up'
# make sure there are machines for non-automated jobs to run
num_free = len(filter(
lambda machine: machine['up'] and machine['locked'] == 0,
machines
))
if num_free < 6 and ctx.owner.startswith('scheduled'):
if ctx.block:
log.info('waiting for more machines to be free...')
time.sleep(10)
continue
else:
assert 0, 'not enough machines free'
newly_locked = lock.lock_many(ctx, config, ctx.owner)
if len(newly_locked) == config:
ctx.config['targets'] = newly_locked
log.info('\n '.join(['Locked targets:', ] + yaml.safe_dump(ctx.config['targets'], default_flow_style=False).splitlines()))
break
elif not ctx.block:
assert 0, 'not enough machines are available'
log.warn('Could not lock enough machines, waiting...')
time.sleep(10)
try:
yield
finally:
if ctx.summary.get('success', False):
log.info('Unlocking machines...')
for machine in ctx.config['targets'].iterkeys():
lock.unlock(ctx, machine, ctx.owner)
def nuke(ctx,
log,
should_unlock,
sync_clocks=True,
reboot_all=True,
noipmi=False):
from teuthology.parallel import parallel
from teuthology.lock import list_locks
total_unnuked = {}
targets = dict(ctx.config['targets'])
if ctx.name:
log.info('Checking targets against current locks')
locks = list_locks(ctx)
#Remove targets who's description doesn't match archive name.
for lock in locks:
for target in targets:
if target == lock['name']:
if ctx.name not in lock['description']:
del ctx.config['targets'][lock['name']]
log.info(
'Not nuking %s because description doesn\'t match',
lock['name'])
with parallel() as p:
for target, hostkey in ctx.config['targets'].iteritems():
p.spawn(
nuke_one,
ctx,
{target: hostkey},
log,
should_unlock,
sync_clocks,
reboot_all,
ctx.config.get('check-locks', True),
noipmi,
)
for unnuked in p:
if unnuked:
total_unnuked.update(unnuked)
if total_unnuked:
log.error('Could not nuke the following targets:\n' + '\n '.join([
'targets:',
] + yaml.safe_dump(total_unnuked,
default_flow_style=False).splitlines()))
def lock_machines(ctx, config):
"""
Lock machines. Called when the teuthology run finds and locks
new machines. This is not called if the one has teuthology-locked
machines and placed those keys in the Targets section of a yaml file.
"""
# It's OK for os_type and os_version to be None here. If we're trying
# to lock a bare metal machine, we'll take whatever is available. If
# we want a vps, defaults will be provided by misc.get_distro and
# misc.get_distro_version in provision.create_if_vm
os_type = ctx.config.get("os_type")
os_version = ctx.config.get("os_version")
arch = ctx.config.get('arch')
log.info('Locking machines...')
assert isinstance(config[0], int), 'config[0] must be an integer'
machine_type = config[1]
how_many = config[0]
# We want to make sure there are always this many machines available
to_reserve = teuth_config.reserve_machines
assert isinstance(to_reserve, int), 'reserve_machines must be integer'
assert (to_reserve >= 0), 'reserve_machines should >= 0'
# change the status during the locking process
report.try_push_job_info(ctx.config, dict(status='waiting'))
while True:
# get a candidate list of machines
machines = lock.list_locks(machine_type=machine_type, up=True,
locked=False, count=how_many + to_reserve)
if machines is None:
if ctx.block:
log.error('Error listing machines, trying again')
time.sleep(20)
continue
else:
raise RuntimeError('Error listing machines')
# make sure there are machines for non-automated jobs to run
if len(machines) < to_reserve + how_many and ctx.owner.startswith('scheduled'):
if ctx.block:
log.info(
'waiting for more machines to be free (need %s + %s, have %s)...',
to_reserve,
how_many,
len(machines),
)
time.sleep(10)
continue
else:
assert 0, ('not enough machines free; need %s + %s, have %s' %
(to_reserve, how_many, len(machines)))
newly_locked = lock.lock_many(ctx, how_many, machine_type, ctx.owner,
ctx.archive, os_type, os_version, arch)
if not newly_locked and not isinstance(newly_locked, list):
raise RuntimeError('Invalid parameters specified')
if len(newly_locked) == how_many:
vmlist = []
for lmach in newly_locked:
if misc.is_vm(lmach):
vmlist.append(lmach)
if vmlist:
log.info('Waiting for virtual machines to come up')
keys_dict = dict()
loopcount = 0
while len(keys_dict) != len(vmlist):
loopcount += 1
time.sleep(10)
keys_dict = lock.ssh_keyscan(vmlist)
log.info('virtual machine is still unavailable')
if loopcount == 40:
loopcount = 0
log.info('virtual machine(s) still not up, ' +
'recreating unresponsive ones.')
for guest in vmlist:
if guest not in keys_dict.keys():
log.info('recreating: ' + guest)
full_name = misc.canonicalize_hostname(guest)
provision.destroy_if_vm(ctx, full_name)
provision.create_if_vm(ctx, full_name)
if lock.do_update_keys(keys_dict):
log.info("Error in virtual machine keys")
newscandict = {}
for dkey in newly_locked.iterkeys():
stats = lockstatus.get_status(dkey)
newscandict[dkey] = stats['ssh_pub_key']
ctx.config['targets'] = newscandict
else:
ctx.config['targets'] = newly_locked
locked_targets = yaml.safe_dump(
ctx.config['targets'],
default_flow_style=False
).splitlines()
log.info('\n '.join(['Locked targets:', ] + locked_targets))
# successfully locked machines, change status back to running
report.try_push_job_info(ctx.config, dict(status='running'))
break
elif not ctx.block:
assert 0, 'not enough machines are available'
else:
how_many = how_many - len(newly_locked)
assert how_many > 0, "lock_machines: how_many counter went" \
"negative, this shouldn't happen"
log.warn('Could not lock enough machines, waiting...')
time.sleep(10)
try:
yield
finally:
# If both unlock_on_failure and nuke-on-error are set, don't unlock now
# because we're just going to nuke (and unlock) later.
unlock_on_failure = (
ctx.config.get('unlock_on_failure', False)
and not ctx.config.get('nuke-on-error', False)
)
if get_status(ctx.summary) == 'pass' or unlock_on_failure:
log.info('Unlocking machines...')
for machine in ctx.config['targets'].iterkeys():
lock.unlock_one(ctx, machine, ctx.owner, ctx.archive)
def lock_machines(ctx, config):
"""
Lock machines. Called when the teuthology run finds and locks
new machines. This is not called if the one has teuthology-locked
machines and placed those keys in the Targets section of a yaml file.
"""
# It's OK for os_type and os_version to be None here. If we're trying
# to lock a bare metal machine, we'll take whatever is available. If
# we want a vps, defaults will be provided by misc.get_distro and
# misc.get_distro_version in provision.create_if_vm
os_type = ctx.config.get("os_type")
os_version = ctx.config.get("os_version")
arch = ctx.config.get('arch')
log.info('Locking machines...')
assert isinstance(config[0], int), 'config[0] must be an integer'
machine_type = config[1]
total_requested = config[0]
# We want to make sure there are always this many machines available
reserved = teuth_config.reserve_machines
assert isinstance(reserved, int), 'reserve_machines must be integer'
assert (reserved >= 0), 'reserve_machines should >= 0'
# change the status during the locking process
report.try_push_job_info(ctx.config, dict(status='waiting'))
all_locked = dict()
requested = total_requested
while True:
# get a candidate list of machines
machines = lock.list_locks(machine_type=machine_type, up=True,
locked=False, count=requested + reserved)
if machines is None:
if ctx.block:
log.error('Error listing machines, trying again')
time.sleep(20)
continue
else:
raise RuntimeError('Error listing machines')
# make sure there are machines for non-automated jobs to run
if len(machines) < reserved + requested and ctx.owner.startswith('scheduled'):
if ctx.block:
log.info(
'waiting for more %s machines to be free (need %s + %s, have %s)...',
machine_type,
reserved,
requested,
len(machines),
)
time.sleep(10)
continue
else:
assert 0, ('not enough machines free; need %s + %s, have %s' %
(reserved, requested, len(machines)))
newly_locked = lock.lock_many(ctx, requested, machine_type, ctx.owner,
ctx.archive, os_type, os_version, arch)
all_locked.update(newly_locked)
log.info(
'{newly_locked} {mtype} machines locked this try, '
'{total_locked}/{total_requested} locked so far'.format(
newly_locked=len(newly_locked),
mtype=machine_type,
total_locked=len(all_locked),
total_requested=total_requested,
)
)
if len(all_locked) == total_requested:
vmlist = []
for lmach in all_locked:
if misc.is_vm(lmach):
vmlist.append(lmach)
if vmlist:
log.info('Waiting for virtual machines to come up')
keys_dict = dict()
loopcount = 0
while len(keys_dict) != len(vmlist):
loopcount += 1
time.sleep(10)
keys_dict = misc.ssh_keyscan(vmlist)
log.info('virtual machine is still unavailable')
if loopcount == 40:
loopcount = 0
log.info('virtual machine(s) still not up, ' +
'recreating unresponsive ones.')
for guest in vmlist:
if guest not in keys_dict.keys():
log.info('recreating: ' + guest)
full_name = misc.canonicalize_hostname(guest)
provision.destroy_if_vm(ctx, full_name)
provision.create_if_vm(ctx, full_name)
if lock.do_update_keys(keys_dict):
log.info("Error in virtual machine keys")
newscandict = {}
for dkey in all_locked.iterkeys():
stats = lockstatus.get_status(dkey)
newscandict[dkey] = stats['ssh_pub_key']
ctx.config['targets'] = newscandict
else:
ctx.config['targets'] = all_locked
locked_targets = yaml.safe_dump(
ctx.config['targets'],
default_flow_style=False
).splitlines()
log.info('\n '.join(['Locked targets:', ] + locked_targets))
# successfully locked machines, change status back to running
report.try_push_job_info(ctx.config, dict(status='running'))
break
elif not ctx.block:
assert 0, 'not enough machines are available'
else:
requested = requested - len(newly_locked)
assert requested > 0, "lock_machines: requested counter went" \
"negative, this shouldn't happen"
log.info(
"{total} machines locked ({new} new); need {more} more".format(
total=len(all_locked), new=len(newly_locked), more=requested)
)
log.warn('Could not lock enough machines, waiting...')
time.sleep(10)
try:
yield
finally:
# If both unlock_on_failure and nuke-on-error are set, don't unlock now
# because we're just going to nuke (and unlock) later.
unlock_on_failure = (
ctx.config.get('unlock_on_failure', False)
and not ctx.config.get('nuke-on-error', False)
)
if get_status(ctx.summary) == 'pass' or unlock_on_failure:
log.info('Unlocking machines...')
for machine in ctx.config['targets'].iterkeys():
lock.unlock_one(ctx, machine, ctx.owner, ctx.archive)
def lock_machines(ctx, config):
log.info('Locking machines...')
assert isinstance(config[0], int), 'config must be an integer'
machine_type = config[1]
config = config[0]
while True:
# make sure there are enough machines up
machines = lock.list_locks(ctx)
if machines is None:
if ctx.block:
log.warn('error listing machines, trying again')
time.sleep(20)
continue
else:
assert 0, 'error listing machines'
num_up = len(
filter(
lambda machine: machine['up'] and machine['type'] ==
machine_type, machines))
assert num_up >= config, 'not enough machines are up'
# make sure there are machines for non-automated jobs to run
num_free = len(
filter(
lambda machine: machine['up'] and machine['locked'] == 0 and
machine['type'] == machine_type, machines))
if num_free < 6 and ctx.owner.startswith('scheduled'):
if ctx.block:
log.info('waiting for more machines to be free...')
time.sleep(10)
continue
else:
assert 0, 'not enough machines free'
newly_locked = lock.lock_many(ctx, config, machine_type, ctx.owner,
ctx.archive)
if len(newly_locked) == config:
vmlist = []
for lmach in newly_locked:
if lock.create_if_vm(ctx, lmach):
vmlist.append(lmach)
if vmlist:
log.info('Waiting for virtual machines to come up')
keyscan_out = ''
loopcount = 0
while len(keyscan_out.splitlines()) != len(vmlist):
loopcount += 1
time.sleep(10)
keyscan_out, current_locks = lock.keyscan_check(
ctx, vmlist)
log.info('virtual machine is stil unavailable')
if loopcount == 40:
loopcount = 0
log.info(
'virtual machine(s) still not up, recreating unresponsive ones.'
)
for guest in vmlist:
if guest not in keyscan_out:
log.info('recreating: ' + guest)
lock.destroy_if_vm(ctx, 'ubuntu@' + guest)
lock.create_if_vm(ctx, 'ubuntu@' + guest)
if lock.update_keys(ctx, keyscan_out, current_locks):
log.info("Error in virtual machine keys")
newscandict = {}
for dkey in newly_locked.iterkeys():
stats = lockstatus.get_status(ctx, dkey)
newscandict[dkey] = stats['sshpubkey']
ctx.config['targets'] = newscandict
else:
ctx.config['targets'] = newly_locked
log.info('\n '.join([
'Locked targets:',
] + yaml.safe_dump(ctx.config['targets'],
default_flow_style=False).splitlines()))
break
elif not ctx.block:
assert 0, 'not enough machines are available'
log.warn('Could not lock enough machines, waiting...')
time.sleep(10)
try:
yield
finally:
if ctx.summary.get('success', False):
log.info('Unlocking machines...')
for machine in ctx.config['targets'].iterkeys():
lock.unlock(ctx, machine, ctx.owner)
def lock_machines(ctx, config):
"""
Lock machines. Called when the teuthology run finds and locks
new machines. This is not called if the one has teuthology-locked
machines and placed those keys in the Targets section of a yaml file.
"""
log.info('Locking machines...')
assert isinstance(config[0], int), 'config[0] must be an integer'
machine_type = config[1]
how_many = config[0]
# We want to make sure there are always this many machines available
to_reserve = 5
while True:
# get a candidate list of machines
machines = lock.list_locks(machine_type=machine_type, up=True,
locked=False, count=how_many + to_reserve)
if machines is None:
if ctx.block:
log.error('Error listing machines, trying again')
time.sleep(20)
continue
else:
raise RuntimeError('Error listing machines')
# make sure there are machines for non-automated jobs to run
if len(machines) <= to_reserve and ctx.owner.startswith('scheduled'):
if ctx.block:
log.info(
'waiting for more machines to be free (need %s see %s)...',
how_many,
len(machines),
)
time.sleep(10)
continue
else:
assert 0, 'not enough machines free'
newly_locked = lock.lock_many(ctx, how_many, machine_type, ctx.owner,
ctx.archive)
if not newly_locked and not isinstance(newly_locked, list):
raise RuntimeError('Invalid parameters specified')
if len(newly_locked) == how_many:
vmlist = []
for lmach in newly_locked:
if misc.is_vm(lmach):
vmlist.append(lmach)
if vmlist:
log.info('Waiting for virtual machines to come up')
keys_dict = dict()
loopcount = 0
while len(keys_dict) != len(vmlist):
loopcount += 1
time.sleep(10)
keys_dict = lock.ssh_keyscan(vmlist)
log.info('virtual machine is still unavailable')
if loopcount == 40:
loopcount = 0
log.info('virtual machine(s) still not up, ' +
'recreating unresponsive ones.')
for guest in vmlist:
if guest not in keys_dict.keys():
log.info('recreating: ' + guest)
full_name = misc.canonicalize_hostname(guest)
provision.destroy_if_vm(ctx, full_name)
provision.create_if_vm(ctx, full_name)
if lock.do_update_keys(keys_dict):
log.info("Error in virtual machine keys")
newscandict = {}
for dkey in newly_locked.iterkeys():
stats = lockstatus.get_status(dkey)
newscandict[dkey] = stats['ssh_pub_key']
ctx.config['targets'] = newscandict
else:
ctx.config['targets'] = newly_locked
# FIXME: Ugh.
log.info('\n '.join(['Locked targets:', ] + yaml.safe_dump(ctx.config['targets'], default_flow_style=False).splitlines()))
break
elif not ctx.block:
assert 0, 'not enough machines are available'
log.warn('Could not lock enough machines, waiting...')
time.sleep(10)
try:
yield
finally:
if ctx.config.get('unlock_on_failure', False) or \
ctx.summary.get('success', False):
log.info('Unlocking machines...')
for machine in ctx.config['targets'].iterkeys():
lock.unlock_one(ctx, machine, ctx.owner)
def lock_machines(ctx, config):
log.info('Locking machines...')
assert isinstance(config[0], int), 'config must be an integer'
machine_type = config[1]
config = config[0]
while True:
# make sure there are enough machines up
machines = lock.list_locks(ctx)
if machines is None:
if ctx.block:
log.warn('error listing machines, trying again')
time.sleep(20)
continue
else:
assert 0, 'error listing machines'
num_up = len(filter(lambda machine: machine['up'] and machine['type'] == machine_type, machines))
assert num_up >= config, 'not enough machines are up'
# make sure there are machines for non-automated jobs to run
num_free = len(filter(
lambda machine: machine['up'] and machine['locked'] == 0 and machine['type'] == machine_type,
machines
))
if num_free < 6 and ctx.owner.startswith('scheduled'):
if ctx.block:
log.info('waiting for more machines to be free...')
time.sleep(10)
continue
else:
assert 0, 'not enough machines free'
newly_locked = lock.lock_many(ctx, config, machine_type, ctx.owner, ctx.archive)
if len(newly_locked) == config:
vmlist = []
for lmach in newly_locked:
if lock.create_if_vm(ctx,lmach):
vmlist.append(lmach)
if vmlist:
log.info('Waiting for virtual machines to come up')
keyscan_out = ''
while len(keyscan_out.splitlines()) != len(vmlist):
time.sleep(10)
keyscan_out, current_locks = lock.keyscan_check(ctx, vmlist)
log.info('virtual machine is stil unavailable')
if lock.update_keys(ctx, keyscan_out, current_locks):
log.info("Error in virtual machine keys")
newscandict = {}
for dkey in newly_locked.iterkeys():
stats = lockstatus.get_status(ctx, dkey)
newscandict[dkey] = stats['sshpubkey']
ctx.config['targets'] = newscandict
else:
ctx.config['targets'] = newly_locked
log.info('\n '.join(['Locked targets:', ] + yaml.safe_dump(ctx.config['targets'], default_flow_style=False).splitlines()))
break
elif not ctx.block:
assert 0, 'not enough machines are available'
log.warn('Could not lock enough machines, waiting...')
time.sleep(10)
try:
yield
finally:
if ctx.summary.get('success', False):
log.info('Unlocking machines...')
for machine in ctx.config['targets'].iterkeys():
lock.unlock(ctx, machine, ctx.owner)
def lock_machines(ctx, config):
log.info("Locking machines...")
assert isinstance(config[0], int), "config[0] must be an integer"
machine_type = config[1]
how_many = config[0]
while True:
# make sure there are enough machines up
machines = lock.list_locks()
if machines is None:
if ctx.block:
log.warn("error listing machines, trying again")
time.sleep(20)
continue
else:
assert 0, "error listing machines"
is_up = lambda machine: machine["up"] and machine["type"] == machine_type # noqa
num_up = len(filter(is_up, machines))
assert num_up >= how_many, "not enough machines are up"
# make sure there are machines for non-automated jobs to run
is_up_and_free = (
lambda machine: machine["up"] and machine["locked"] == 0 and machine["type"] == machine_type
) # noqa
up_and_free = filter(is_up_and_free, machines)
num_free = len(up_and_free)
if num_free < 6 and ctx.owner.startswith("scheduled"):
if ctx.block:
log.info("waiting for more machines to be free (need %s see %s)...", how_many, num_free)
time.sleep(10)
continue
else:
assert 0, "not enough machines free"
newly_locked = lock.lock_many(ctx, how_many, machine_type, ctx.owner, ctx.archive)
if len(newly_locked) == how_many:
vmlist = []
for lmach in newly_locked:
if lock.create_if_vm(ctx, lmach):
vmlist.append(lmach)
if vmlist:
log.info("Waiting for virtual machines to come up")
keyscan_out = ""
loopcount = 0
while len(keyscan_out.splitlines()) != len(vmlist):
loopcount += 1
time.sleep(10)
keyscan_out, current_locks = lock.keyscan_check(ctx, vmlist)
log.info("virtual machine is stil unavailable")
if loopcount == 40:
loopcount = 0
log.info("virtual machine(s) still not up, " + "recreating unresponsive ones.")
for guest in vmlist:
if guest not in keyscan_out:
log.info("recreating: " + guest)
lock.destroy_if_vm(ctx, "ubuntu@" + guest)
lock.create_if_vm(ctx, "ubuntu@" + guest)
if lock.update_keys(ctx, keyscan_out, current_locks):
log.info("Error in virtual machine keys")
newscandict = {}
for dkey in newly_locked.iterkeys():
stats = lockstatus.get_status(ctx, dkey)
newscandict[dkey] = stats["sshpubkey"]
ctx.config["targets"] = newscandict
else:
ctx.config["targets"] = newly_locked
# FIXME: Ugh.
log.info(
"\n ".join(
["Locked targets:"] + yaml.safe_dump(ctx.config["targets"], default_flow_style=False).splitlines()
)
)
break
elif not ctx.block:
assert 0, "not enough machines are available"
log.warn("Could not lock enough machines, waiting...")
time.sleep(10)
try:
yield
finally:
if ctx.summary.get("success", False):
log.info("Unlocking machines...")
for machine in ctx.config["targets"].iterkeys():
lock.unlock_one(ctx, machine, ctx.owner)
def lock_machines(ctx, config):
"""
Lock machines. Called when the teuthology run finds and locks
new machines. This is not called if the one has teuthology-locked
machines and placed those keys in the Targets section of a yaml file.
"""
log.info('Locking machines...')
assert isinstance(config[0], int), 'config[0] must be an integer'
machine_type = config[1]
machine_types = teuthology.get_multi_machine_types(machine_type)
how_many = config[0]
while True:
# make sure there are enough machines up
machines = lock.list_locks()
if machines is None:
if ctx.block:
log.warn('error listing machines, trying again')
time.sleep(20)
continue
else:
assert 0, 'error listing machines'
is_up = lambda machine: machine['up'] and machine['type'] in machine_types # noqa
num_up = len(filter(is_up, machines))
assert num_up >= how_many, 'not enough machines are up'
# make sure there are machines for non-automated jobs to run
is_up_and_free = lambda machine: machine['up'] and machine['locked'] == 0 and machine['type'] in machine_types # noqa
up_and_free = filter(is_up_and_free, machines)
num_free = len(up_and_free)
if num_free < 6 and ctx.owner.startswith('scheduled'):
if ctx.block:
log.info(
'waiting for more machines to be free (need %s see %s)...',
how_many,
num_free,
)
time.sleep(10)
continue
else:
assert 0, 'not enough machines free'
newly_locked = lock.lock_many(ctx, how_many, machine_type, ctx.owner,
ctx.archive)
if len(newly_locked) == how_many:
vmlist = []
for lmach in newly_locked:
if teuthology.is_vm(lmach):
vmlist.append(lmach)
if vmlist:
log.info('Waiting for virtual machines to come up')
keyscan_out = ''
loopcount = 0
while len(keyscan_out.splitlines()) != len(vmlist):
loopcount += 1
time.sleep(10)
keyscan_out, current_locks = lock.keyscan_check(ctx,
vmlist)
log.info('virtual machine is still unavailable')
if loopcount == 40:
loopcount = 0
log.info('virtual machine(s) still not up, ' +
'recreating unresponsive ones.')
for guest in vmlist:
if guest not in keyscan_out:
log.info('recreating: ' + guest)
lock.destroy_if_vm(ctx, 'ubuntu@' + guest)
lock.create_if_vm(ctx, 'ubuntu@' + guest)
if lock.update_keys(ctx, keyscan_out, current_locks):
log.info("Error in virtual machine keys")
newscandict = {}
for dkey in newly_locked.iterkeys():
stats = lockstatus.get_status(ctx, dkey)
newscandict[dkey] = stats['sshpubkey']
ctx.config['targets'] = newscandict
else:
ctx.config['targets'] = newly_locked
# FIXME: Ugh.
log.info('\n '.join(['Locked targets:', ] + yaml.safe_dump(ctx.config['targets'], default_flow_style=False).splitlines()))
break
elif not ctx.block:
assert 0, 'not enough machines are available'
log.warn('Could not lock enough machines, waiting...')
time.sleep(10)
try:
yield
finally:
if ctx.config.get('unlock_on_failure', False) or \
ctx.summary.get('success', False):
log.info('Unlocking machines...')
for machine in ctx.config['targets'].iterkeys():
lock.unlock_one(ctx, machine, ctx.owner)
def lock_machines(ctx, config):
"""
Lock machines. Called when the teuthology run finds and locks
new machines. This is not called if the one has teuthology-locked
machines and placed those keys in the Targets section of a yaml file.
"""
log.info('Locking machines...')
assert isinstance(config[0], int), 'config[0] must be an integer'
machine_type = config[1]
machine_types = teuthology.get_multi_machine_types(machine_type)
how_many = config[0]
while True:
# make sure there are enough machines up
machines = lock.list_locks()
if machines is None:
if ctx.block:
log.warn('error listing machines, trying again')
time.sleep(20)
continue
else:
assert 0, 'error listing machines'
is_up = lambda machine: machine['up'] and machine['type'] in machine_types # noqa
num_up = len(filter(is_up, machines))
assert num_up >= how_many, 'not enough machines are up'
# make sure there are machines for non-automated jobs to run
is_up_and_free = lambda machine: machine['up'] and machine['locked'] == 0 and machine['type'] in machine_types # noqa
up_and_free = filter(is_up_and_free, machines)
num_free = len(up_and_free)
if num_free < 6 and ctx.owner.startswith('scheduled'):
if ctx.block:
log.info(
'waiting for more machines to be free (need %s see %s)...',
how_many,
num_free,
)
time.sleep(10)
continue
else:
assert 0, 'not enough machines free'
newly_locked = lock.lock_many(ctx, how_many, machine_type, ctx.owner,
ctx.archive)
if len(newly_locked) == how_many:
vmlist = []
for lmach in newly_locked:
if teuthology.is_vm(lmach):
vmlist.append(lmach)
if vmlist:
log.info('Waiting for virtual machines to come up')
keyscan_out = ''
loopcount = 0
while len(keyscan_out.splitlines()) != len(vmlist):
loopcount += 1
time.sleep(10)
keyscan_out, current_locks = lock.keyscan_check(ctx,
vmlist)
log.info('virtual machine is still unavailable')
if loopcount == 40:
loopcount = 0
log.info('virtual machine(s) still not up, ' +
'recreating unresponsive ones.')
for guest in vmlist:
if guest not in keyscan_out:
log.info('recreating: ' + guest)
lock.destroy_if_vm(ctx, 'ubuntu@' + guest)
lock.create_if_vm(ctx, 'ubuntu@' + guest)
if lock.update_keys(ctx, keyscan_out, current_locks):
log.info("Error in virtual machine keys")
newscandict = {}
for dkey in newly_locked.iterkeys():
stats = lockstatus.get_status(ctx, dkey)
newscandict[dkey] = stats['sshpubkey']
ctx.config['targets'] = newscandict
else:
ctx.config['targets'] = newly_locked
# FIXME: Ugh.
log.info('\n '.join(['Locked targets:', ] + yaml.safe_dump(ctx.config['targets'], default_flow_style=False).splitlines()))
break
elif not ctx.block:
assert 0, 'not enough machines are available'
log.warn('Could not lock enough machines, waiting...')
time.sleep(10)
try:
yield
finally:
if ctx.config.get('unlock_on_failure', False) or \
ctx.summary.get('success', False):
log.info('Unlocking machines...')
for machine in ctx.config['targets'].iterkeys():
lock.unlock_one(ctx, machine, ctx.owner)
def lock_machines(ctx, config):
"""
Lock machines. Called when the teuthology run finds and locks
new machines. This is not called if the one has teuthology-locked
machines and placed those keys in the Targets section of a yaml file.
"""
# It's OK for os_type and os_version to be None here. If we're trying
# to lock a bare metal machine, we'll take whatever is available. If
# we want a vps, defaults will be provided by misc.get_distro and
# misc.get_distro_version in provision.create_if_vm
os_type = ctx.config.get("os_type")
os_version = ctx.config.get("os_version")
arch = ctx.config.get("arch")
log.info("Locking machines...")
assert isinstance(config[0], int), "config[0] must be an integer"
machine_type = config[1]
total_requested = config[0]
# We want to make sure there are always this many machines available
reserved = teuth_config.reserve_machines
assert isinstance(reserved, int), "reserve_machines must be integer"
assert reserved >= 0, "reserve_machines should >= 0"
# change the status during the locking process
report.try_push_job_info(ctx.config, dict(status="waiting"))
all_locked = dict()
requested = total_requested
while True:
# get a candidate list of machines
machines = lock.list_locks(machine_type=machine_type, up=True, locked=False, count=requested + reserved)
if machines is None:
if ctx.block:
log.error("Error listing machines, trying again")
time.sleep(20)
continue
else:
raise RuntimeError("Error listing machines")
# make sure there are machines for non-automated jobs to run
if len(machines) < reserved + requested and ctx.owner.startswith("scheduled"):
if ctx.block:
log.info(
"waiting for more %s machines to be free (need %s + %s, have %s)...",
machine_type,
reserved,
requested,
len(machines),
)
time.sleep(10)
continue
else:
assert 0, "not enough machines free; need %s + %s, have %s" % (reserved, requested, len(machines))
newly_locked = lock.lock_many(ctx, requested, machine_type, ctx.owner, ctx.archive, os_type, os_version, arch)
all_locked.update(newly_locked)
log.info(
"{newly_locked} {mtype} machines locked this try, "
"{total_locked}/{total_requested} locked so far".format(
newly_locked=len(newly_locked),
mtype=machine_type,
total_locked=len(all_locked),
total_requested=total_requested,
)
)
if len(all_locked) == total_requested:
vmlist = []
for lmach in all_locked:
if misc.is_vm(lmach):
vmlist.append(lmach)
if vmlist:
log.info("Waiting for virtual machines to come up")
keys_dict = dict()
loopcount = 0
while len(keys_dict) != len(vmlist):
loopcount += 1
time.sleep(10)
keys_dict = misc.ssh_keyscan(vmlist)
log.info("virtual machine is still unavailable")
if loopcount == 40:
loopcount = 0
log.info("virtual machine(s) still not up, " + "recreating unresponsive ones.")
for guest in vmlist:
if guest not in keys_dict.keys():
log.info("recreating: " + guest)
full_name = misc.canonicalize_hostname(guest)
provision.destroy_if_vm(ctx, full_name)
provision.create_if_vm(ctx, full_name)
if lock.do_update_keys(keys_dict):
log.info("Error in virtual machine keys")
newscandict = {}
for dkey in all_locked.iterkeys():
stats = lockstatus.get_status(dkey)
newscandict[dkey] = stats["ssh_pub_key"]
ctx.config["targets"] = newscandict
else:
ctx.config["targets"] = all_locked
locked_targets = yaml.safe_dump(ctx.config["targets"], default_flow_style=False).splitlines()
log.info("\n ".join(["Locked targets:"] + locked_targets))
# successfully locked machines, change status back to running
report.try_push_job_info(ctx.config, dict(status="running"))
break
elif not ctx.block:
assert 0, "not enough machines are available"
else:
requested = requested - len(newly_locked)
assert requested > 0, "lock_machines: requested counter went" "negative, this shouldn't happen"
log.info(
"{total} machines locked ({new} new); need {more} more".format(
total=len(all_locked), new=len(newly_locked), more=requested
)
)
log.warn("Could not lock enough machines, waiting...")
time.sleep(10)
try:
yield
finally:
# If both unlock_on_failure and nuke-on-error are set, don't unlock now
# because we're just going to nuke (and unlock) later.
unlock_on_failure = ctx.config.get("unlock_on_failure", False) and not ctx.config.get("nuke-on-error", False)
if get_status(ctx.summary) == "pass" or unlock_on_failure:
log.info("Unlocking machines...")
for machine in ctx.config["targets"].iterkeys():
lock.unlock_one(ctx, machine, ctx.owner, ctx.archive)
def lock_machines(ctx, config):
"""
Lock machines. Called when the teuthology run finds and locks
new machines. This is not called if the one has teuthology-locked
machines and placed those keys in the Targets section of a yaml file.
"""
# It's OK for os_type and os_version to be None here. If we're trying
# to lock a bare metal machine, we'll take whatever is available. If
# we want a vps, defaults will be provided by misc.get_distro and
# misc.get_distro_version in provision.create_if_vm
os_type = ctx.config.get("os_type")
os_version = ctx.config.get("os_version")
arch = ctx.config.get('arch')
log.info('Locking machines...')
assert isinstance(config[0], int), 'config[0] must be an integer'
machine_type = config[1]
how_many = config[0]
# We want to make sure there are always this many machines available
to_reserve = 5
# change the status during the locking process
report.try_push_job_info(ctx.config, dict(status='waiting'))
while True:
# get a candidate list of machines
machines = lock.list_locks(machine_type=machine_type,
up=True,
locked=False,
count=how_many + to_reserve)
if machines is None:
if ctx.block:
log.error('Error listing machines, trying again')
time.sleep(20)
continue
else:
raise RuntimeError('Error listing machines')
# make sure there are machines for non-automated jobs to run
if len(machines) < to_reserve + how_many and ctx.owner.startswith(
'scheduled'):
if ctx.block:
log.info(
'waiting for more machines to be free (need %s + %s, have %s)...',
to_reserve,
how_many,
len(machines),
)
time.sleep(10)
continue
else:
assert 0, ('not enough machines free; need %s + %s, have %s' %
(to_reserve, how_many, len(machines)))
newly_locked = lock.lock_many(ctx, how_many, machine_type, ctx.owner,
ctx.archive, os_type, os_version, arch)
if not newly_locked and not isinstance(newly_locked, list):
raise RuntimeError('Invalid parameters specified')
if len(newly_locked) == how_many:
vmlist = []
for lmach in newly_locked:
if misc.is_vm(lmach):
vmlist.append(lmach)
if vmlist:
log.info('Waiting for virtual machines to come up')
keys_dict = dict()
loopcount = 0
while len(keys_dict) != len(vmlist):
loopcount += 1
time.sleep(10)
keys_dict = lock.ssh_keyscan(vmlist)
log.info('virtual machine is still unavailable')
if loopcount == 40:
loopcount = 0
log.info('virtual machine(s) still not up, ' +
'recreating unresponsive ones.')
for guest in vmlist:
if guest not in keys_dict.keys():
log.info('recreating: ' + guest)
full_name = misc.canonicalize_hostname(guest)
provision.destroy_if_vm(ctx, full_name)
provision.create_if_vm(ctx, full_name)
if lock.do_update_keys(keys_dict):
log.info("Error in virtual machine keys")
newscandict = {}
for dkey in newly_locked.iterkeys():
stats = lockstatus.get_status(dkey)
newscandict[dkey] = stats['ssh_pub_key']
ctx.config['targets'] = newscandict
else:
ctx.config['targets'] = newly_locked
locked_targets = yaml.safe_dump(
ctx.config['targets'], default_flow_style=False).splitlines()
log.info('\n '.join([
'Locked targets:',
] + locked_targets))
# successfully locked machines, change status back to running
report.try_push_job_info(ctx.config, dict(status='running'))
break
elif not ctx.block:
assert 0, 'not enough machines are available'
log.warn('Could not lock enough machines, waiting...')
time.sleep(10)
try:
yield
finally:
if ctx.config.get('unlock_on_failure', False) or \
get_status(ctx.summary) == 'pass':
log.info('Unlocking machines...')
for machine in ctx.config['targets'].iterkeys():
lock.unlock_one(ctx, machine, ctx.owner, ctx.archive)
