def on_event(data, message, dry_run, treeherder_server_url, acknowledge, **kwargs):
if ignored(data):
if acknowledge:
message.ack()
return 0 # SUCCESS
# Cleaning mozci caches
buildjson.BUILDS_CACHE = {}
query_jobs.JOBS_CACHE = {}
repo_name = data["project"]
action = data["action"]
times = data["times"]
# Pulse gives us resultset_id, we need to get revision from it.
resultset_id = data["resultset_id"]
treeherder_client = TreeherderClient(server_url=treeherder_server_url)
LOG.info("%s action requested by %s on repo_name %s with resultset_id: %s" % (
data['action'],
data["requester"],
data["project"],
data["resultset_id"])
)
revision = treeherder_client.get_resultsets(repo_name, id=resultset_id)[0]["revision"]
status = None
if action == "trigger_missing_jobs":
mgr = BuildAPIManager()
mgr.trigger_missing_jobs_for_revision(repo_name, revision, dry_run=dry_run)
if acknowledge:
status = 'trigger_missing_jobs request sent'
else:
status = 'Dry-mode, no request sent'
elif action == "trigger_all_talos_jobs":
trigger_all_talos_jobs(
repo_name=repo_name,
revision=revision,
times=times,
priority=-1,
dry_run=dry_run
)
if acknowledge:
status = 'trigger_all_talos_jobs: {0} times request sent with priority'\
'lower then normal'.format(times)
else:
status = 'Dry-mode, no request sent'
else:
raise Exception(
'We were not aware of the "{}" action. Please address the code.'.format(action)
)
LOG.debug(status)
if acknowledge:
# We need to ack the message to remove it from our queue
message.ack()
return 0 # SUCCESS
def on_resultset_action_event(data, message, dry_run):
# Cleaning mozci caches
buildjson.BUILDS_CACHE = {}
query_jobs.JOBS_CACHE = {}
repo_name = data["project"]
action = data["action"]
times = data["times"]
# Pulse gives us resultset_id, we need to get revision from it.
resultset_id = data["resultset_id"]
treeherder_client = TreeherderClient()
# We do not handle 'cancel_all' action right now, so skip it.
if action == "cancel_all":
message.ack()
return
LOG.info("%s action requested by %s on repo_name %s with resultset_id: %s" %
(data['action'], data["requester"], data["project"], data["resultset_id"]))
revision = treeherder_client.get_resultsets(repo_name, id=resultset_id)[0]["revision"]
status = None
if action == "trigger_missing_jobs":
trigger_missing_jobs_for_revision(repo_name, revision, dry_run=dry_run)
if not dry_run:
status = 'trigger_missing_jobs request sent'
else:
status = 'Dry-mode, no request sent'
elif action == "trigger_all_talos_jobs":
trigger_all_talos_jobs(repo_name, revision, times, dry_run=dry_run)
if not dry_run:
status = 'trigger_all_talos_jobs %s times request sent' % times
else:
status = 'Dry-mode, no request sent'
# Send a pulse message showing what we did
message_sender = MessageHandler()
pulse_message = {
'resultset_id': resultset_id,
'action': action,
'requester': data['requester'],
'status': status}
routing_key = '{}.{}'.format(repo_name, action)
message_sender.publish_message(pulse_message, routing_key)
# We need to ack the message to remove it from our queue
message.ack()
def on_event(data, message, dry_run, treeherder_server_url, **kwargs):
if ignored(data):
return 0 # SUCCESS
# Cleaning mozci caches
buildjson.BUILDS_CACHE = {}
query_jobs.JOBS_CACHE = {}
repo_name = data["project"]
action = data["action"]
times = data["times"]
# Pulse gives us resultset_id, we need to get revision from it.
resultset_id = data["resultset_id"]
treeherder_client = TreeherderClient(server_url=treeherder_server_url)
LOG.info(
"%s action requested by %s on repo_name %s with resultset_id: %s" %
(data['action'], data["requester"], data["project"],
data["resultset_id"]))
revision = treeherder_client.get_resultsets(repo_name,
id=resultset_id)[0]["revision"]
if action == "trigger_missing_jobs":
mgr = BuildAPIManager()
mgr.trigger_missing_jobs_for_revision(repo_name,
revision,
dry_run=dry_run)
elif action == "trigger_all_talos_jobs":
trigger_all_talos_jobs(repo_name=repo_name,
revision=revision,
times=times,
priority=-1,
dry_run=dry_run)
else:
raise Exception(
'We were not aware of the "{}" action. Please address the code.'.
format(action))
return 0 # SUCCESS
def main():
alerts = getAlerts()
for alert in alerts:
# new alert
LOG.info("Running alert for: [%s, %s, %s]" % (alert['test'], alert['buildername'], alert['revision']))
if alert['stage'] == 0:
LOG.info("We are in stage 0.")
if checkMerge(alert['revision'], alert['buildername']) or 'pgo' in alert['buildername']:
LOG.info("We are ignoring alert: %s since it is either a merge or a pgo job." % alert['test'])
alert['stage'] = -1 # We need to have manual inspection in this case.
alert['user'] = '******'
updateAlert(alert['id'], alert['revision'], alert['buildername'], alert['test'],
alert['stage'], alert['loop'], alert['user'])
else:
alert['stage'] = 1
# trigger jobs for backfill
if alert['stage'] == 1:
LOG.info("We are in stage 1, and going to backfill jobs.")
revisionList = getRevisions(alert['revision'], alert['buildername'], start=-2, end=2)
# Setting Treeherder as the source for querying.
set_query_source("treeherder")
trigger_range(alert['buildername'], revisionList, times=6, dry_run=DRY_RUN)
alert['stage'] = 2
# We want some time interval between stage 1 and 2, so we exit.
updateAlert(alert['id'], alert['revision'], alert['buildername'], alert['test'],
alert['stage'], alert['loop'], alert['user'])
continue
# verify jobs for backfill
if alert['stage'] == 2:
LOG.info("We are in stage 2, and going to verify if jobs are backfilled.")
revisionList = getRevisions(alert['revision'], alert['buildername'], start=-2, end=2)
for revision in revisionList:
dataPoints = getSuccessfulJobs(revision, alert['buildername'])
# If dataPoints are less than 6, it means that builds/jobs are still running.
if dataPoints < 6:
print "data points <6 for revision: %s" % revision
# We wait for 6 hours for all triggered tests to complete,
# And if they don't then we mark them for manual intervention/
alert['loop'] += 1
if alert['loop'] > (TIME_TO_BUILD + TIME_TO_TEST + PENDING_TIME) / CYCLE_TIME:
LOG.info("The jobs did not complete backfilling in time, assigning for human inspection.")
alert['stage'] = -1
alert['user'] = '******'
else:
LOG.info("The jobs have not completed backfilling. Looping back to stage 1.")
alert['stage'] = 1
break
if alert['stage'] != 2:
print "updating alert and then continue, not stage 2"
updateAlert(alert['id'], alert['revision'], alert['buildername'], alert['test'],
alert['stage'], alert['loop'], alert['user'])
continue
badRevisions = []
# Reset the loop for upcoming stages
alert['loop'] = 0
for i in range(1, len(revisionList)):
print "getting results for revision number: %s" % i
results = compare(alert['test'], alert['buildername'], revisionList[i], revisionList[i-1])
print "compare returned: %s" % results
if results < -2.0:
print "appending bad revision to list: %s"% revisionList[i]
badRevisions.append(revisionList[i])
if len(badRevisions) != 1:
LOG.info("There are too many bad revisions: %s for alert %s on buildername %s, "
"assigning for human inspection." % (badRevisions, alert['test'], alert['buildername']))
alert['stage'] = -1 # too noisy, something bad happened
alert['user'] = '******'
print "too many bad revisions, update alert to human"
updateAlert(alert['id'], alert['revision'], alert['buildername'], alert['test'],
alert['stage'], alert['loop'], alert['user'])
continue
if checkMerge(badRevisions[0], alert['buildername']):
LOG.info("The bad revision %s identified for alert %s on buildername %s is a merge, "
"assigning for human inspection" % (badRevisions[0], alert['test'], alert['buildername']))
alert['stage'] = -1 # A merge revision is a bad revision, manually inspect
alert['user'] = '******'
if alert['revision'] != badRevisions[0]:
LOG.info("Alert_Manager misreported the bad revision. The actual bad revision is %s "
"for alert %s on %s buildername." % (badRevisions[0], alert['test'], alert['buildername']))
alert['revision'] = badRevisions[0] # we misreported initially, change the actual regression revision
print "setting stage = 3!"
alert['stage'] = 3
# Trigger all talos stage
if alert['stage'] == 3:
LOG.info("We are in stage 3, and going to trigger all_talos jobs.")
repo_name = query_repo_name_from_buildername(alert['buildername'])
# Setting Treeherder as the source for querying.
set_query_source("treeherder")
trigger_all_talos_jobs(repo_name, alert['revision'], times=6, dry_run=DRY_RUN)
previousRevision = getRevisions(alert['revision'], alert['buildername'], start=-1, end=-1)[0]
trigger_all_talos_jobs(repo_name, previousRevision, times=6, dry_run=DRY_RUN)
alert['stage'] = 4
updateAlert(alert['id'], alert['revision'], alert['buildername'], alert['test'],
alert['stage'], alert['loop'], alert['user'])
continue
# Verify All talos stage is completed
if alert['stage'] == 4:
LOG.info("We are in stage 4, and going to verify if all_talos ran successfully.")
previousRevision = getRevisions(alert['revision'], alert['buildername'], start=-1, end=-1)[0]
repo_name = query_repo_name_from_buildername(alert['buildername'])
all_buildernames = build_talos_buildernames_for_repo(repo_name)
for revision in [alert['revision'], previousRevision]:
for buildername in all_buildernames:
dataPoints = getSuccessfulJobs(revision, buildername)
if dataPoints < 6:
# We wait for 8 hours for all talos tests to complete,
# And if they don't then we mark them for manual intervention
alert['loop'] += 1
if alert['loop'] > (TIME_TO_BUILD + TIME_TO_TEST + PENDING_TIME + TIME_TO_WAIT) / CYCLE_TIME:
LOG.info("The all talos jobs for alert %s on %s revision did not complete in time, "
" assigning for human inspection." % (alert['test'], alert['revision']))
alert['stage'] = -1
alert['user'] = '******'
else:
alert['stage'] = 3
break
if alert['stage'] != 4:
break
if alert['stage'] != 4:
updateAlert(alert['id'], alert['revision'], alert['buildername'], alert['test'],
alert['stage'], alert['loop'], alert['user'])
continue
alert['stage'] = 5 # final stage, sheriff will check for this.
alert['user'] = '******'
LOG.info("All automated parts are complete.")
updateAlert(alert['id'], alert['revision'], alert['buildername'], alert['test'], alert['stage'], alert['loop'], alert['user'])
def main():
options = parse_args()
if options.debug:
LOG = setup_logging(logging.DEBUG)
else:
LOG = setup_logging(logging.INFO)
if options.action == 'trigger-all-talos':
trigger_all_talos_jobs(options.repo_name,
options.rev,
options.times,
dry_run=options.dry_run)
sys.exit(0)
validate_options(options)
if not options.dry_run and not valid_credentials():
sys.exit(-1)
# Setting the QUERY_SOURCE global variable in mozci.py
set_query_source(options.query_source)
if options.buildernames:
options.buildernames = sanitize_buildernames(options.buildernames)
repo_url = query_repo_url_from_buildername(options.buildernames[0])
if not options.repo_name:
repo_name = query_repo_name_from_buildername(options.buildernames[0])
else:
repo_name = options.repo_name
repo_url = query_repo_url(repo_name)
if options.rev == 'tip':
revision = query_repo_tip(repo_url).changesets[0].node
LOG.info("The tip of %s is %s", repo_name, revision)
else:
revision = query_push_by_revision(repo_url,
options.rev,
return_revision_list=True)
# Schedule jobs through TaskCluster if --taskcluster option has been set to true
if options.taskcluster:
mgr = TaskClusterBuildbotManager(web_auth=True)
else:
mgr = BuildAPIManager()
trigger_build_if_missing = options.trigger_build_if_missing
if repo_name == 'try':
trigger_build_if_missing = False
# Mode 0: Backfill
if options.backfill:
manual_backfill(revision,
options.buildernames[0],
dry_run=options.dry_run)
return
# Mode 1: Trigger coalesced jobs
if options.coalesced:
query_api = BuildApi()
request_ids = query_api.find_all_jobs_by_status(
repo_name, revision, COALESCED)
if len(request_ids) == 0:
LOG.info('We did not find any coalesced job')
for request_id in request_ids:
make_retrigger_request(repo_name=repo_name,
request_id=request_id,
auth=get_credentials(),
dry_run=options.dry_run)
return
# Mode #2: Fill-in a revision or trigger_test_jobs_only
if options.fill_revision or options.trigger_tests_only:
mgr.trigger_missing_jobs_for_revision(
repo_name=repo_name,
revision=revision,
dry_run=options.dry_run,
trigger_build_if_missing=not options.trigger_tests_only)
return
# Mode #3: Trigger jobs based on revision list modifiers
if not (options.includes or options.exclude or options.failed_jobs
or options.trigger_talos_for_build):
job_names = options.buildernames
# Mode 4 - Schedule every builder matching --includes and does not match --exclude.
elif options.includes or options.exclude:
_includes_excludes(options)
# Mode 5: Use --failed-jobs to trigger jobs for particular revision
elif options.failed_jobs:
job_names = TreeherderApi().find_all_jobs_by_status(
repo_name=repo_name, revision=revision, status=WARNING)
elif options.trigger_talos_for_build:
trigger_talos_jobs_for_build(
buildername=options.buildernames[0],
revision=revision,
times=2,
dry_run=options.dry_run,
)
exit(0)
for buildername in job_names:
revlist = determine_revlist(repo_url=repo_url,
buildername=buildername,
rev=revision,
back_revisions=options.back_revisions,
delta=options.delta,
from_rev=options.from_rev,
backfill=options.backfill,
skips=options.skips,
max_revisions=options.max_revisions)
_print_treeherder_link(revlist=revlist,
repo_name=repo_name,
buildername=buildername,
revision=revision,
log=LOG,
includes=options.includes,
exclude=options.exclude)
try:
mgr.trigger_range(
buildername=buildername,
repo_name=repo_name,
revisions=revlist,
times=options.times,
dry_run=options.dry_run,
files=options.files,
trigger_build_if_missing=trigger_build_if_missing)
except Exception as e:
LOG.exception(e)
exit(1)
def main():
options = parse_args()
if options.debug:
LOG = setup_logging(logging.DEBUG)
else:
LOG = setup_logging(logging.INFO)
if options.action == 'trigger-all-talos':
trigger_all_talos_jobs(options.repo_name, options.rev, options.times,
dry_run=options.dry_run)
sys.exit(0)
validate_options(options)
if not options.dry_run and not valid_credentials():
sys.exit(-1)
# Setting the QUERY_SOURCE global variable in mozci.py
set_query_source(options.query_source)
if options.buildernames:
options.buildernames = sanitize_buildernames(options.buildernames)
repo_url = query_repo_url_from_buildername(options.buildernames[0])
if not options.repo_name:
repo_name = query_repo_name_from_buildername(options.buildernames[0])
else:
repo_name = options.repo_name
repo_url = query_repo_url(repo_name)
if options.rev == 'tip':
revision = query_repo_tip(repo_url).changesets[0].node
LOG.info("The tip of %s is %s", repo_name, revision)
else:
revision = query_push_by_revision(repo_url, options.rev,
return_revision_list=True)
# Schedule jobs through TaskCluster if --taskcluster option has been set to true
if options.taskcluster:
mgr = TaskClusterBuildbotManager(web_auth=True)
else:
mgr = BuildAPIManager()
trigger_build_if_missing = options.trigger_build_if_missing
if repo_name == 'try':
trigger_build_if_missing = False
# Mode 0: Backfill
if options.backfill:
manual_backfill(revision, options.buildernames[0], dry_run=options.dry_run)
return
# Mode 1: Trigger coalesced jobs
if options.coalesced:
query_api = BuildApi()
request_ids = query_api.find_all_jobs_by_status(repo_name,
revision, COALESCED)
if len(request_ids) == 0:
LOG.info('We did not find any coalesced job')
for request_id in request_ids:
make_retrigger_request(repo_name=repo_name,
request_id=request_id,
auth=get_credentials(),
dry_run=options.dry_run)
return
# Mode #2: Fill-in a revision or trigger_test_jobs_only
if options.fill_revision or options.trigger_tests_only:
mgr.trigger_missing_jobs_for_revision(
repo_name=repo_name,
revision=revision,
dry_run=options.dry_run,
trigger_build_if_missing=not options.trigger_tests_only
)
return
# Mode #3: Trigger jobs based on revision list modifiers
if not (options.includes or options.exclude or options.failed_jobs or
options.trigger_talos_for_build):
job_names = options.buildernames
# Mode 4 - Schedule every builder matching --includes and does not match --exclude.
elif options.includes or options.exclude:
_includes_excludes(options)
# Mode 5: Use --failed-jobs to trigger jobs for particular revision
elif options.failed_jobs:
job_names = TreeherderApi().find_all_jobs_by_status(
repo_name=repo_name,
revision=revision,
status=WARNING)
elif options.trigger_talos_for_build:
trigger_talos_jobs_for_build(
buildername=options.buildernames[0],
revision=revision,
times=2,
dry_run=options.dry_run,
)
exit(0)
for buildername in job_names:
revlist = determine_revlist(
repo_url=repo_url,
buildername=buildername,
rev=revision,
back_revisions=options.back_revisions,
delta=options.delta,
from_rev=options.from_rev,
backfill=options.backfill,
skips=options.skips,
max_revisions=options.max_revisions)
_print_treeherder_link(
revlist=revlist,
repo_name=repo_name,
buildername=buildername,
revision=revision,
log=LOG,
includes=options.includes,
exclude=options.exclude)
try:
mgr.trigger_range(
buildername=buildername,
repo_name=repo_name,
revisions=revlist,
times=options.times,
dry_run=options.dry_run,
files=options.files,
trigger_build_if_missing=trigger_build_if_missing
)
except Exception as e:
LOG.exception(e)
exit(1)
def on_event(data, message, dry_run, treeherder_server_url, acknowledge, **kwargs):
LOG.info('Acknowledge value: {}'.format(acknowledge))
if ignored(data):
if acknowledge:
LOG.info('Message acknowledged')
message.ack()
return 0 # SUCCESS
# Cleaning mozci caches
buildjson.BUILDS_CACHE = {}
query_jobs.JOBS_CACHE = {}
repo_name = data["project"]
action = data["action"]
times = data["times"]
# Pulse gives us resultset_id, we need to get revision from it.
resultset_id = data["resultset_id"]
treeherder_client = TreeherderClient(server_url=treeherder_server_url)
LOG.info("%s action requested by %s on repo_name %s with resultset_id: %s" % (
data['action'],
data["requester"],
data["project"],
data["resultset_id"])
)
revision = treeherder_client.get_resultsets(repo_name, id=resultset_id)[0]["revision"]
status = None
if action == "trigger_missing_jobs":
mgr = BuildAPIManager()
mgr.trigger_missing_jobs_for_revision(repo_name, revision, dry_run=dry_run)
if acknowledge:
status = 'trigger_missing_jobs request sent'
else:
status = 'Dry-mode, no request sent'
elif action == "trigger_all_talos_jobs":
trigger_all_talos_jobs(
repo_name=repo_name,
revision=revision,
times=times,
priority=-1,
dry_run=dry_run
)
if acknowledge:
status = 'trigger_all_talos_jobs: {0} times request sent with priority'\
'lower then normal'.format(times)
else:
status = 'Dry-mode, no request sent'
else:
raise Exception(
'We were not aware of the "{}" action. Please address the code.'.format(action)
)
LOG.debug(status)
if acknowledge:
# We need to ack the message to remove it from our queue
LOG.info('Message acknowledged')
message.ack()
return 0 # SUCCESS
def main():
alerts = getAlerts()
for alert in alerts:
# new alert
if alert['stage'] == 0:
if checkMerge(alert['revision'], alert['buildername']) or 'pgo' in alert['buildername']:
LOG.info("We are ignoring this alert since it is either a merge or a pgo job.")
alert['stage'] = -1 # We need to have manual inspection in this case.
alert['user'] = '******'
updateAlert(alert['id'], alert['revision'], alert['buildername'], alert['test'],
alert['stage'], alert['loop'], alert['user'])
else:
alert['stage'] = 1
# trigger jobs for backfill
if alert['stage'] == 1:
LOG.info("We are in stage 1, and going to backfill jobs.")
revisionList = getRevisions(alert['revision'], alert['buildername'], start=-2, end=2)
trigger_range(alert['buildername'], revisionList, times=6, dry_run=DRY_RUN)
alert['stage'] = 2
# We want some time interval between stage 1 and 2, so we exit.
updateAlert(alert['id'], alert['revision'], alert['buildername'], alert['test'],
alert['stage'], alert['loop'], alert['user'])
continue
# verify jobs for backfill
if alert['stage'] == 2:
LOG.info("We are in stage 2, and going to verify if jobs are backfilled.")
revisionList = getRevisions(alert['revision'], alert['buildername'], start=-2, end=2)
for revision in revisionList:
dataPoints = getSuccessfulJobs(revision, alert['buildername'])
# If dataPoints are less than 6, it means that builds/jobs are still running.
if dataPoints < 6:
# We wait for 6 hours for all triggered tests to complete,
# And if they don't then we mark them for manual intervention/
alert['loop'] += 1
if alert['loop'] > (TIME_TO_BUILD + TIME_TO_TEST + PENDING_TIME) / CYCLE_TIME:
alert['stage'] = -1
alert['user'] = '******'
else:
alert['stage'] = 1
break
if alert['stage'] != 2:
updateAlert(alert['id'], alert['revision'], alert['buildername'], alert['test'],
alert['stage'], alert['loop'], alert['user'])
continue
badRevisions = []
# Reset the loop for upcoming stages
alert['loop'] = 0
for i in range(1, len(revisionList)):
results = compare(alert['test'], alert['buildername'], revisionList[i], revisionList[i-1])
if results < -2.0:
badRevisions.append(revisionList[i])
if len(badRevisions) != 1:
alert['stage'] = -1 # too noisy, something bad happened
alert['user'] = '******'
updateAlert(alert['id'], alert['revision'], alert['buildername'], alert['test'],
alert['stage'], alert['loop'], alert['user'])
continue
if checkMerge(badRevisions[0], alert['buildername']):
alert['stage'] = -1 # A merge revision is a bad revision, manually inspect
alert['user'] = '******'
if alert['revision'] != badRevisions[0]:
alert['revision'] = badRevisions[0] # we misreported initially, change the actual regression revision
alert['stage'] = 3
# Trigger all talos stage
if alert['stage'] == 3:
LOG.info("We are in stage 3, and going to trigger all_talos jobs.")
repo_name = query_repo_name_from_buildername(alert['buildername'])
trigger_all_talos_jobs(repo_name, alert['revision'], times=6, dry_run=DRY_RUN)
previousRevision = getRevisions(alert['revision'], alert['buildername'], start=-1, end=-1)[0]
trigger_all_talos_jobs(repo_name, previousRevision, times=6, dry_run=DRY_RUN)
alert['stage'] = 4
updateAlert(alert['id'], alert['revision'], alert['buildername'], alert['test'],
alert['stage'], alert['loop'], alert['user'])
continue
# Verify All talos stage is completed
if alert['stage'] == 4:
LOG.info("We are in stage 4, and going to verify if all_talos ran successfully.")
previousRevision = getRevisions(alert['revision'], alert['buildername'], start=-1, end=-1)[0]
repo_name = query_repo_name_from_buildername(alert['buildername'])
all_buildernames = build_talos_buildernames_for_repo(repo_name)
for revision in [alert['revision'], previousRevision]:
for buildername in all_buildernames:
dataPoints = getSuccessfulJobs(revision, buildername)
if dataPoints < 6:
# We wait for 8 hours for all talos tests to complete,
# And if they don't then we mark them for manual intervention
alert['loop'] += 1
if alert['loop'] > (TIME_TO_BUILD + TIME_TO_TEST + PENDING_TIME + TIME_TO_WAIT) / CYCLE_TIME:
alert['stage'] = -1
alert['user'] = '******'
else:
alert['stage'] = 3
break
if alert['stage'] != 4:
break
if alert['stage'] != 4:
updateAlert(alert['id'], alert['revision'], alert['buildername'], alert['test'],
alert['stage'], alert['loop'], alert['user'])
continue
alert['stage'] = 5 # final stage, sheriff will check for this.
updateAlert(alert['id'], alert['revision'], alert['buildername'], alert['test'], alert['stage'], alert['loop'], alert['user'])
def main():
options = parse_args()
if options.debug:
LOG = setup_logging(logging.DEBUG)
else:
LOG = setup_logging(logging.INFO)
if options.action == 'trigger-all-talos':
trigger_all_talos_jobs(options.repo_name, options.rev, options.times,
dry_run=options.dry_run)
sys.exit(0)
validate_options(options)
if not options.dry_run and not valid_credentials():
sys.exit(-1)
# Setting the QUERY_SOURCE global variable in mozci.py
set_query_source(options.query_source)
if options.buildernames:
options.buildernames = sanitize_buildernames(options.buildernames)
repo_url = query_repo_url_from_buildername(options.buildernames[0])
if not options.repo_name:
repo_name = query_repo_name_from_buildername(options.buildernames[0])
else:
repo_name = options.repo_name
repo_url = query_repo_url(repo_name)
if options.rev == 'tip':
revision = query_repo_tip(repo_url).changesets[0].node
LOG.info("The tip of %s is %s", repo_name, revision)
else:
revision = query_push_by_revision(repo_url, options.rev,
return_revision_list=True)
# Schedule jobs through TaskCluster if --taskcluster option has been set to true
if options.taskcluster:
mgr = TaskClusterBuildbotManager()
else:
mgr = BuildAPIManager()
trigger_build_if_missing = options.trigger_build_if_missing
if repo_name == 'try':
trigger_build_if_missing = False
# Mode 0: Backfill
if options.backfill:
manual_backfill(revision, options.buildernames[0], dry_run=options.dry_run)
return
# Mode 1: Trigger coalesced jobs
if options.coalesced:
query_api = BuildApi()
request_ids = query_api.find_all_jobs_by_status(repo_name,
revision, COALESCED)
if len(request_ids) == 0:
LOG.info('We did not find any coalesced job')
for request_id in request_ids:
make_retrigger_request(repo_name=repo_name,
request_id=request_id,
auth=get_credentials(),
dry_run=options.dry_run)
return
# Mode #2: Fill-in a revision or trigger_test_jobs_only
if options.fill_revision or options.trigger_tests_only:
mgr.trigger_missing_jobs_for_revision(
repo_name=repo_name,
revision=revision,
dry_run=options.dry_run,
trigger_build_if_missing=not options.trigger_tests_only
)
return
# Mode #3: Trigger jobs based on revision list modifiers
if not (options.includes or options.exclude or options.failed_jobs or
options.trigger_talos_for_build):
job_names = options.buildernames
# Mode 4 - Schedule every builder matching --includes and does not match --exclude.
elif options.includes or options.exclude:
filters_in = options.includes.split(',') + [repo_name]
filters_out = []
if options.exclude:
filters_out = options.exclude.split(',')
job_names = filter_buildernames(
buildernames=query_builders(repo_name=repo_name),
include=filters_in,
exclude=filters_out
)
if len(job_names) == 0:
LOG.info("0 jobs match these filters. please try again.")
return
if options.existing_only:
# We query all successful jobs for a given revision and filter
# them by include/exclude filters.
trigger_build_if_missing = False
successful_jobs = TreeherderApi().find_all_jobs_by_status(
repo_name=repo_name,
revision=revision,
status=SUCCESS)
# We will filter out all the existing job from those successful job we have.
job_names = [buildername for buildername in successful_jobs
if buildername in job_names]
cont = raw_input("The ones which have existing builds out of %i jobs will be triggered,\
do you wish to continue? y/n/d (d=show details) " % len(job_names))
else:
cont = raw_input("%i jobs will be triggered, do you wish to continue? \
y/n/d (d=show details) " % len(job_names))
if cont.lower() == 'd':
LOG.info("The following jobs will be triggered: \n %s" % '\n'.join(job_names))
cont = raw_input("Do you wish to continue? y/n ")
if cont.lower() != 'y':
exit(1)
# Mode 5: Use --failed-jobs to trigger jobs for particular revision
elif options.failed_jobs:
job_names = TreeherderApi().find_all_jobs_by_status(
repo_name=repo_name,
revision=revision,
status=WARNING)
elif options.trigger_talos_for_build:
trigger_talos_jobs_for_build(
buildername=options.buildernames[0],
revision=revision,
times=2,
dry_run=options.dry_run,
)
exit(0)
for buildername in job_names:
revlist = determine_revlist(
repo_url=repo_url,
buildername=buildername,
rev=revision,
back_revisions=options.back_revisions,
delta=options.delta,
from_rev=options.from_rev,
backfill=options.backfill,
skips=options.skips,
max_revisions=options.max_revisions)
_print_treeherder_link(
revlist=revlist,
repo_name=repo_name,
buildername=buildername,
revision=revision,
log=LOG,
includes=options.includes,
exclude=options.exclude)
try:
mgr.trigger_range(
buildername=buildername,
repo_name=repo_name,
revisions=revlist,
times=options.times,
dry_run=options.dry_run,
files=options.files,
trigger_build_if_missing=trigger_build_if_missing
)
except Exception, e:
LOG.exception(e)
exit(1)
def main():
alerts = getAlerts()
for alert in alerts:
# new alert
if alert['stage'] == 0:
if checkMerge(
alert['revision'],
alert['buildername']) or 'pgo' in alert['buildername']:
LOG.info(
"We are ignoring this alert since it is either a merge or a pgo job."
)
alert[
'stage'] = -1 # We need to have manual inspection in this case.
alert['user'] = '******'
updateAlert(alert['id'], alert['revision'],
alert['buildername'], alert['test'],
alert['stage'], alert['loop'], alert['user'])
else:
alert['stage'] = 1
# trigger jobs for backfill
if alert['stage'] == 1:
LOG.info("We are in stage 1, and going to backfill jobs.")
revisionList = getRevisions(alert['revision'],
alert['buildername'],
start=-2,
end=2)
trigger_range(alert['buildername'],
revisionList,
times=6,
dry_run=DRY_RUN)
alert['stage'] = 2
# We want some time interval between stage 1 and 2, so we exit.
updateAlert(alert['id'], alert['revision'], alert['buildername'],
alert['test'], alert['stage'], alert['loop'],
alert['user'])
continue
# verify jobs for backfill
if alert['stage'] == 2:
LOG.info(
"We are in stage 2, and going to verify if jobs are backfilled."
)
revisionList = getRevisions(alert['revision'],
alert['buildername'],
start=-2,
end=2)
for revision in revisionList:
dataPoints = getSuccessfulJobs(revision, alert['buildername'])
# If dataPoints are less than 6, it means that builds/jobs are still running.
if dataPoints < 6:
# We wait for 6 hours for all triggered tests to complete,
# And if they don't then we mark them for manual intervention/
alert['loop'] += 1
if alert['loop'] > (TIME_TO_BUILD + TIME_TO_TEST +
PENDING_TIME) / CYCLE_TIME:
alert['stage'] = -1
alert['user'] = '******'
else:
alert['stage'] = 1
break
if alert['stage'] != 2:
updateAlert(alert['id'], alert['revision'],
alert['buildername'], alert['test'],
alert['stage'], alert['loop'], alert['user'])
continue
badRevisions = []
# Reset the loop for upcoming stages
alert['loop'] = 0
for i in range(1, len(revisionList)):
results = compare(alert['test'], alert['buildername'],
revisionList[i], revisionList[i - 1])
if results < -2.0:
badRevisions.append(revisionList[i])
if len(badRevisions) != 1:
alert['stage'] = -1 # too noisy, something bad happened
alert['user'] = '******'
updateAlert(alert['id'], alert['revision'],
alert['buildername'], alert['test'],
alert['stage'], alert['loop'], alert['user'])
continue
if checkMerge(badRevisions[0], alert['buildername']):
alert[
'stage'] = -1 # A merge revision is a bad revision, manually inspect
alert['user'] = '******'
if alert['revision'] != badRevisions[0]:
alert['revision'] = badRevisions[
0] # we misreported initially, change the actual regression revision
alert['stage'] = 3
# Trigger all talos stage
if alert['stage'] == 3:
LOG.info("We are in stage 3, and going to trigger all_talos jobs.")
repo_name = query_repo_name_from_buildername(alert['buildername'])
trigger_all_talos_jobs(repo_name,
alert['revision'],
times=6,
dry_run=DRY_RUN)
previousRevision = getRevisions(alert['revision'],
alert['buildername'],
start=-1,
end=-1)[0]
trigger_all_talos_jobs(repo_name,
previousRevision,
times=6,
dry_run=DRY_RUN)
alert['stage'] = 4
updateAlert(alert['id'], alert['revision'], alert['buildername'],
alert['test'], alert['stage'], alert['loop'],
alert['user'])
continue
# Verify All talos stage is completed
if alert['stage'] == 4:
LOG.info(
"We are in stage 4, and going to verify if all_talos ran successfully."
)
previousRevision = getRevisions(alert['revision'],
alert['buildername'],
start=-1,
end=-1)[0]
repo_name = query_repo_name_from_buildername(alert['buildername'])
all_buildernames = build_talos_buildernames_for_repo(repo_name)
for revision in [alert['revision'], previousRevision]:
for buildername in all_buildernames:
dataPoints = getSuccessfulJobs(revision, buildername)
if dataPoints < 6:
# We wait for 8 hours for all talos tests to complete,
# And if they don't then we mark them for manual intervention
alert['loop'] += 1
if alert['loop'] > (TIME_TO_BUILD + TIME_TO_TEST +
PENDING_TIME +
TIME_TO_WAIT) / CYCLE_TIME:
alert['stage'] = -1
alert['user'] = '******'
else:
alert['stage'] = 3
break
if alert['stage'] != 4:
break
if alert['stage'] != 4:
updateAlert(alert['id'], alert['revision'],
alert['buildername'], alert['test'],
alert['stage'], alert['loop'], alert['user'])
continue
alert['stage'] = 5 # final stage, sheriff will check for this.
updateAlert(alert['id'], alert['revision'], alert['buildername'],
alert['test'], alert['stage'], alert['loop'],
alert['user'])
def on_resultset_action_event(data, message, dry_run, stage=False):
# Cleaning mozci caches
buildjson.BUILDS_CACHE = {}
query_jobs.JOBS_CACHE = {}
repo_name = data["project"]
action = data["action"]
times = data["times"]
# Pulse gives us resultset_id, we need to get revision from it.
resultset_id = data["resultset_id"]
if stage:
treeherder_client = TreeherderClient(host='treeherder.allizom.org')
else:
treeherder_client = TreeherderClient()
# We do not handle 'cancel_all' action right now, so skip it.
if action == "cancel_all":
if not dry_run:
message.ack()
return
LOG.info("%s action requested by %s on repo_name %s with resultset_id: %s" % (
data['action'],
data["requester"],
data["project"],
data["resultset_id"])
)
revision = treeherder_client.get_resultsets(repo_name, id=resultset_id)[0]["revision"]
status = None
if action == "trigger_missing_jobs":
mgr = BuildAPIManager()
mgr.trigger_missing_jobs_for_revision(repo_name, revision, dry_run=dry_run)
if not dry_run:
status = 'trigger_missing_jobs request sent'
else:
status = 'Dry-mode, no request sent'
elif action == "trigger_all_talos_jobs":
trigger_all_talos_jobs(
repo_name=repo_name,
revision=revision,
times=times,
priority=-1,
dry_run=dry_run
)
if not dry_run:
status = 'trigger_all_talos_jobs: {0} times request sent with priority'\
'lower then normal'.format(times)
else:
status = 'Dry-mode, no request sent'
# Send a pulse message showing what we did
message_sender = MessageHandler()
pulse_message = {
'resultset_id': resultset_id,
'action': action,
'requester': data['requester'],
'status': status}
routing_key = '{}.{}'.format(repo_name, action)
try:
message_sender.publish_message(pulse_message, routing_key)
except:
LOG.warning("Failed to publish message over pulse stream.")
if not dry_run:
# We need to ack the message to remove it from our queue
message.ack()
