def sync_phase(phase):
phase_steps = list(phase.steps)
if phase.date_started is None:
phase.date_started = safe_agg(min, (s.date_started for s in phase_steps))
db.session.add(phase)
if phase_steps:
if all(s.status == Status.finished for s in phase_steps):
phase.status = Status.finished
phase.date_finished = safe_agg(max, (s.date_finished for s in phase_steps))
else:
# ensure we dont set the status to finished unless it actually is
new_status = aggregate_status((s.status for s in phase_steps))
if new_status != Status.finished:
phase.status = new_status
if any(s.result is Result.failed for s in phase_steps):
phase.result = Result.failed
elif phase.status == Status.finished:
phase.result = aggregate_result((s.result for s in phase_steps))
if db.session.is_modified(phase):
phase.date_modified = datetime.utcnow()
db.session.add(phase)
db.session.commit()
def sync_phase(phase):
phase_steps = list(phase.steps)
if phase.date_started is None:
phase.date_started = safe_agg(
min, (s.date_started for s in phase_steps), phase.date_started)
db.session.add(phase)
if all(s.status == Status.finished for s in phase_steps):
phase.status = Status.finished
phase.date_finished = safe_agg(
max, (s.date_finished for s in phase_steps), phase.date_finished)
if any(s.result is Result.failed for s in phase_steps):
phase.result = Result.failed
else:
phase.result = safe_agg(
max, (s.result for s in phase.steps), Result.unknown)
db.session.add(phase)
elif any(s.status != Status.finished for s in phase_steps):
phase.status = Status.in_progress
db.session.add(phase)
if db.session.is_modified(phase):
phase.date_modified = datetime.utcnow()
db.session.commit()
def sync_phase(phase):
phase_steps = list(phase.steps)
if phase.date_started is None:
phase.date_started = safe_agg(min,
(s.date_started for s in phase_steps))
db.session.add(phase)
if phase_steps:
if all(s.status == Status.finished for s in phase_steps):
phase.status = Status.finished
phase.date_finished = safe_agg(max, (s.date_finished
for s in phase_steps))
else:
# ensure we dont set the status to finished unless it actually is
new_status = aggregate_status((s.status for s in phase_steps))
if new_status != Status.finished:
phase.status = new_status
if any(s.result is Result.failed for s in phase_steps):
phase.result = Result.failed
elif phase.status == Status.finished:
phase.result = aggregate_result((s.result for s in phase_steps))
if db.session.is_modified(phase):
phase.date_modified = datetime.utcnow()
db.session.add(phase)
db.session.commit()
def sync_phase(phase):
phase_steps = list(phase.steps)
if phase.date_started is None:
phase.date_started = safe_agg(min,
(s.date_started for s in phase_steps))
db.session.add(phase)
if phase_steps:
if all(s.status == Status.finished for s in phase_steps):
phase.status = Status.finished
phase.date_finished = safe_agg(max, (s.date_finished
for s in phase_steps))
elif any(s.status != Status.finished for s in phase_steps):
phase.status = Status.in_progress
if any(s.result is Result.failed for s in phase_steps):
phase.result = Result.failed
elif phase.status == Status.finished:
phase.result = safe_agg(max, (s.result for s in phase.steps))
if db.session.is_modified(phase):
phase.date_modified = datetime.utcnow()
db.session.add(phase)
db.session.commit()
def sync_phase(phase, implementation):
_find_and_retry_jobsteps(phase, implementation)
phase_steps = list(phase.steps)
if phase.date_started is None:
phase.date_started = safe_agg(min,
(s.date_started for s in phase_steps))
db.session.add(phase)
if phase_steps:
if all(s.status == Status.finished for s in phase_steps):
phase.status = Status.finished
phase.date_finished = safe_agg(max, (s.date_finished
for s in phase_steps))
else:
# ensure we dont set the status to finished unless it actually is
new_status = aggregate_status((s.status for s in phase_steps))
if new_status != Status.finished:
phase.status = new_status
if any(s.result is Result.failed for s in phase_steps):
phase.result = Result.failed
if phase.status == Status.finished:
# Sets the final phase result.
implementation.validate_phase(phase=phase)
if db.session.is_modified(phase):
phase.date_modified = datetime.utcnow()
db.session.add(phase)
db.session.commit()
def sync_phase(phase, implementation):
_find_and_retry_jobsteps(phase, implementation)
phase_steps = list(phase.steps)
if phase.date_started is None:
phase.date_started = safe_agg(min, (s.date_started for s in phase_steps))
db.session.add(phase)
if phase_steps:
if all(s.status == Status.finished for s in phase_steps):
phase.status = Status.finished
phase.date_finished = safe_agg(max, (s.date_finished for s in phase_steps))
else:
# ensure we dont set the status to finished unless it actually is
new_status = aggregate_status((s.status for s in phase_steps))
if new_status != Status.finished:
phase.status = new_status
if any(s.result is Result.failed for s in phase_steps):
phase.result = Result.failed
if phase.status == Status.finished:
# Sets the final phase result.
implementation.validate_phase(phase=phase)
if db.session.is_modified(phase):
phase.date_modified = datetime.utcnow()
db.session.add(phase)
db.session.commit()
def sync_build(build_id):
"""
Synchronizing the build happens continuously until all jobs have reported in
as finished or have failed/aborted.
This task is responsible for:
- Checking in with jobs
- Aborting/retrying them if they're beyond limits
- Aggregating the results from jobs into the build itself
"""
build = Build.query.get(build_id)
if not build:
return
if build.status == Status.finished:
return
all_jobs = list(Job.query.filter(
Job.build_id == build_id,
))
is_finished = sync_build.verify_all_children() == Status.finished
if any(p.status != Status.finished for p in all_jobs):
is_finished = False
prev_started = build.date_started
build.date_started = safe_agg(
min, (j.date_started for j in all_jobs if j.date_started))
# We want to report how long we waited for the build to start once and only once,
# so we do it at the transition from not started to started.
if not prev_started and build.date_started:
queued_time = build.date_started - build.date_created
statsreporter.stats().log_timing('build_start_latency', _timedelta_to_millis(queued_time))
if is_finished:
# If there are no jobs (or no jobs with a finished date) fall back to
# finishing now, since at this point, the build is done executing.
build.date_finished = safe_agg(
max, (j.date_finished for j in all_jobs if j.date_finished), datetime.utcnow())
else:
build.date_finished = None
if build.date_started and build.date_finished:
build.duration = _timedelta_to_millis(build.date_finished - build.date_started)
else:
build.duration = None
if any(j.result is Result.failed for j in all_jobs):
build.result = Result.failed
elif is_finished:
build.result = aggregate_result((j.result for j in all_jobs))
else:
build.result = Result.unknown
if is_finished:
build.status = Status.finished
else:
# ensure we dont set the status to finished unless it actually is
new_status = aggregate_status((j.status for j in all_jobs))
if new_status != Status.finished:
build.status = new_status
if is_finished:
build.date_decided = datetime.utcnow()
decided_latency = build.date_decided - build.date_finished
statsreporter.stats().log_timing('build_decided_latency', _timedelta_to_millis(decided_latency))
else:
build.date_decided = None
if db.session.is_modified(build):
build.date_modified = datetime.utcnow()
db.session.add(build)
db.session.commit()
if not is_finished:
raise sync_build.NotFinished
with statsreporter.stats().timer('build_stat_aggregation'):
try:
aggregate_build_stat(build, 'test_count')
aggregate_build_stat(build, 'test_duration')
aggregate_build_stat(build, 'test_failures')
aggregate_build_stat(build, 'test_rerun_count')
aggregate_build_stat(build, 'tests_missing')
aggregate_build_stat(build, 'lines_covered')
aggregate_build_stat(build, 'lines_uncovered')
aggregate_build_stat(build, 'diff_lines_covered')
aggregate_build_stat(build, 'diff_lines_uncovered')
except Exception:
current_app.logger.exception('Failing recording aggregate stats for build %s', build.id)
fire_signal.delay(
signal='build.finished',
kwargs={'build_id': build.id.hex},
)
queue.delay('update_project_stats', kwargs={
'project_id': build.project_id.hex,
}, countdown=1)
def sync_job(job_id):
job = Job.query.get(job_id)
if not job:
return
if job.status == Status.finished:
return
# TODO(dcramer): we make an assumption that there is a single step
job_plan = JobPlan.query.options(subqueryload_all('plan.steps')).filter(
JobPlan.job_id == job.id, ).join(Plan).first()
try:
if not job_plan:
raise UnrecoverableException(
'Got sync_job task without job plan: %s' % (job.id, ))
try:
step = job_plan.plan.steps[0]
except IndexError:
raise UnrecoverableException('Missing steps for plan')
implementation = step.get_implementation()
implementation.update(job=job)
except UnrecoverableException:
job.status = Status.finished
job.result = Result.aborted
current_app.logger.exception('Unrecoverable exception syncing %s',
job.id)
is_finished = sync_job.verify_all_children() == Status.finished
if is_finished:
job.status = Status.finished
all_phases = list(job.phases)
job.date_started = safe_agg(min, (j.date_started
for j in all_phases if j.date_started))
if is_finished:
job.date_finished = safe_agg(max,
(j.date_finished
for j in all_phases if j.date_finished))
else:
job.date_finished = None
if job.date_started and job.date_finished:
job.duration = int(
(job.date_finished - job.date_started).total_seconds() * 1000)
else:
job.duration = None
# if any phases are marked as failing, fail the build
if any(j.result is Result.failed for j in all_phases):
job.result = Result.failed
# if any test cases were marked as failing, fail the build
elif TestCase.query.filter(TestCase.result == Result.failed,
TestCase.job_id == job.id).first():
job.result = Result.failed
# if we've finished all phases, use the best result available
elif is_finished:
job.result = safe_agg(max, (j.result for j in all_phases),
Result.unknown)
else:
job.result = Result.unknown
if is_finished:
job.status = Status.finished
elif any(j.status is Status.in_progress for j in all_phases):
job.status = Status.in_progress
else:
job.status = Status.queued
if db.session.is_modified(job):
job.date_modified = datetime.utcnow()
db.session.add(job)
db.session.commit()
publish_job_update(job)
if not is_finished:
raise sync_job.NotFinished
_record_tests_missing(job)
queue.delay('notify_job_finished', kwargs={
'job_id': job.id.hex,
})
if job_plan:
queue.delay('update_project_plan_stats',
kwargs={
'project_id': job.project_id.hex,
'plan_id': job_plan.plan_id.hex,
},
countdown=1)
def _sync_step_from_active(self, step):
try:
job_name = step.data['job_name']
build_no = step.data['build_no']
except KeyError:
raise UnrecoverableException('Missing Jenkins job information')
try:
item = self._get_response('/job/{}/{}'.format(
job_name, build_no))
except NotFound:
raise UnrecoverableException('Unable to find job in Jenkins')
# TODO(dcramer): we're doing a lot of work here when we might
# not need to due to it being sync'd previously
node, _ = get_or_create(Node, where={
'label': item['builtOn'],
})
step.node = node
step.label = item['fullDisplayName']
step.date_started = datetime.utcfromtimestamp(
item['timestamp'] / 1000)
if item['building']:
step.status = Status.in_progress
else:
step.status = Status.finished
step.result = RESULT_MAP[item['result']]
# values['duration'] = item['duration'] or None
step.date_finished = datetime.utcfromtimestamp(
(item['timestamp'] + item['duration']) / 1000)
# step.data.update({
# 'backend': {
# 'uri': item['url'],
# 'label': item['fullDisplayName'],
# }
# })
db.session.add(step)
# TODO(dcramer): we shoudl abstract this into a sync_phase
phase = step.phase
if not phase.date_started:
phase.date_started = safe_agg(
min, (s.date_started for s in phase.steps), step.date_started)
db.session.add(phase)
if phase.status != step.status:
phase.status = step.status
db.session.add(phase)
if step.status == Status.finished:
phase.status = Status.finished
phase.date_finished = safe_agg(
max, (s.date_finished for s in phase.steps), step.date_finished)
if any(s.result is Result.failed for s in phase.steps):
phase.result = Result.failed
else:
phase.result = safe_agg(
max, (s.result for s in phase.steps), Result.unknown)
db.session.add(phase)
db.session.commit()
if step.status != Status.finished:
return
# sync artifacts
for artifact in item.get('artifacts', ()):
artifact, created = get_or_create(Artifact, where={
'step': step,
'name': artifact['fileName'],
}, defaults={
'project': step.project,
'job': step.job,
'data': artifact,
})
sync_artifact.delay_if_needed(
artifact_id=artifact.id.hex,
task_id=artifact.id.hex,
parent_task_id=step.id.hex,
)
# sync test results
try:
with db.session.begin_nested():
self._sync_test_results(
step=step,
job_name=job_name,
build_no=build_no,
)
except Exception:
self.logger.exception(
'Failed to sync test results for %s #%s', job_name, build_no)
else:
db.session.commit()
# sync console log
try:
result = True
while result:
result = self._sync_log(
jobstep=step,
name=step.label,
job_name=job_name,
build_no=build_no,
)
except Exception:
db.session.rollback()
current_app.logger.exception(
'Unable to sync console log for job step %r',
step.id.hex)
def sync_job(job_id):
with RCount('sync_job'):
job = Job.query.get(job_id)
if not job:
return
if job.status == Status.finished:
return
# TODO(dcramer): we make an assumption that there is a single step
jobplan, implementation = JobPlan.get_build_step_for_job(job_id=job.id)
try:
implementation.update(job=job)
except UnrecoverableException:
job.status = Status.finished
job.result = Result.aborted
current_app.logger.exception('Unrecoverable exception syncing %s', job.id)
all_phases = list(job.phases)
# propagate changes to any phases as they live outside of the
# normalize synchronization routines
sync_job_phases(job, all_phases)
is_finished = sync_job.verify_all_children() == Status.finished
if any(p.status != Status.finished for p in all_phases):
is_finished = False
job.date_started = safe_agg(
min, (j.date_started for j in all_phases if j.date_started))
if is_finished:
job.date_finished = safe_agg(
max, (j.date_finished for j in all_phases if j.date_finished))
else:
job.date_finished = None
if job.date_started and job.date_finished:
job.duration = int((job.date_finished - job.date_started).total_seconds() * 1000)
else:
job.duration = None
# if any phases are marked as failing, fail the build
if any(j.result is Result.failed for j in all_phases):
job.result = Result.failed
# if any test cases were marked as failing, fail the build
elif TestCase.query.filter(TestCase.result == Result.failed, TestCase.job_id == job.id).first():
job.result = Result.failed
# if we've finished all phases, use the best result available
elif is_finished:
job.result = aggregate_result((j.result for j in all_phases))
else:
job.result = Result.unknown
if is_finished:
job.status = Status.finished
else:
# ensure we dont set the status to finished unless it actually is
new_status = aggregate_status((j.status for j in all_phases))
if new_status != Status.finished:
job.status = new_status
elif job.status == Status.finished:
job.status = Status.in_progress
current_app.logger.exception('Job incorrectly marked as finished: %s', job.id)
if db.session.is_modified(job):
job.date_modified = datetime.utcnow()
db.session.add(job)
db.session.commit()
if not is_finished:
raise sync_job.NotFinished
try:
aggregate_job_stat(job, 'test_count')
aggregate_job_stat(job, 'test_duration')
aggregate_job_stat(job, 'test_failures')
aggregate_job_stat(job, 'test_rerun_count')
aggregate_job_stat(job, 'tests_missing')
aggregate_job_stat(job, 'lines_covered')
aggregate_job_stat(job, 'lines_uncovered')
aggregate_job_stat(job, 'diff_lines_covered')
aggregate_job_stat(job, 'diff_lines_uncovered')
except Exception:
current_app.logger.exception('Failing recording aggregate stats for job %s', job.id)
fire_signal.delay(
signal='job.finished',
kwargs={'job_id': job.id.hex},
)
if jobplan:
queue.delay('update_project_plan_stats', kwargs={
'project_id': job.project_id.hex,
'plan_id': jobplan.plan_id.hex,
}, countdown=1)
def sync_build(build_id):
"""
Synchronizing the build happens continuously until all jobs have reported in
as finished or have failed/aborted.
This task is responsible for:
- Checking in with jobs
- Aborting/retrying them if they're beyond limits
- Aggregating the results from jobs into the build itself
"""
build = Build.query.get(build_id)
if not build:
return
if build.status == Status.finished:
return
all_jobs = list(Job.query.filter(Job.build_id == build_id, ))
is_finished = sync_build.verify_all_children() == Status.finished
build.date_started = safe_agg(min, (j.date_started
for j in all_jobs if j.date_started))
if is_finished:
build.date_finished = safe_agg(max,
(j.date_finished
for j in all_jobs if j.date_finished))
else:
build.date_finished = None
if build.date_started and build.date_finished:
build.duration = int(
(build.date_finished - build.date_started).total_seconds() * 1000)
else:
build.duration = None
if any(j.result is Result.failed for j in all_jobs):
build.result = Result.failed
elif is_finished:
build.result = safe_agg(max, (j.result for j in all_jobs),
Result.unknown)
else:
build.result = Result.unknown
if is_finished:
build.status = Status.finished
elif any(j.status is Status.in_progress for j in all_jobs):
build.status = Status.in_progress
else:
build.status = Status.queued
if db.session.is_modified(build):
build.date_modified = datetime.utcnow()
db.session.add(build)
db.session.commit()
publish_build_update(build)
if not is_finished:
raise sync_build.NotFinished
_record_tests_missing(build)
queue.delay('notify_build_finished', kwargs={
'build_id': build.id.hex,
})
queue.delay('update_project_stats',
kwargs={
'project_id': build.project_id.hex,
},
countdown=1)
def sync_build(build_id):
"""
Synchronizing the build happens continuously until all jobs have reported in
as finished or have failed/aborted.
This task is responsible for:
- Checking in with jobs
- Aborting/retrying them if they're beyond limits
- Aggregating the results from jobs into the build itself
"""
build = Build.query.get(build_id)
if not build:
return
if build.status == Status.finished:
return
all_jobs = list(Job.query.filter(
Job.build_id == build_id,
))
is_finished = sync_build.verify_all_children() == Status.finished
build.date_started = safe_agg(
min, (j.date_started for j in all_jobs if j.date_started))
if is_finished:
build.date_finished = safe_agg(
max, (j.date_finished for j in all_jobs if j.date_finished))
else:
build.date_finished = None
if build.date_started and build.date_finished:
build.duration = int((build.date_finished - build.date_started).total_seconds() * 1000)
else:
build.duration = None
if any(j.result is Result.failed for j in all_jobs):
build.result = Result.failed
elif is_finished:
build.result = safe_agg(
max, (j.result for j in all_jobs), Result.unknown)
else:
build.result = Result.unknown
if is_finished:
build.status = Status.finished
elif any(j.status is Status.in_progress for j in all_jobs):
build.status = Status.in_progress
else:
build.status = Status.queued
if db.session.is_modified(build):
build.date_modified = datetime.utcnow()
db.session.add(build)
db.session.commit()
publish_build_update(build)
if not is_finished:
raise sync_build.NotFinished
_record_tests_missing(build)
queue.delay('notify_build_finished', kwargs={
'build_id': build.id.hex,
})
queue.delay('update_project_stats', kwargs={
'project_id': build.project_id.hex,
}, countdown=1)
def sync_job(job_id):
"""
Updates jobphase and job statuses based on the status of the constituent jobsteps.
"""
job = Job.query.get(job_id)
if not job:
return
if job.status == Status.finished:
return
# TODO(dcramer): we make an assumption that there is a single step
jobplan, implementation = JobPlan.get_build_step_for_job(job_id=job.id)
try:
implementation.update(job=job)
except UnrecoverableException:
job.status = Status.finished
job.result = Result.infra_failed
current_app.logger.exception('Unrecoverable exception syncing %s',
job.id)
all_phases = list(job.phases)
# propagate changes to any phases as they live outside of the
# normalize synchronization routines
sync_job_phases(job, all_phases, implementation)
is_finished = sync_job.verify_all_children() == Status.finished
if any(p.status != Status.finished for p in all_phases):
is_finished = False
job.date_started = safe_agg(min, (j.date_started
for j in all_phases if j.date_started))
if is_finished:
job.date_finished = safe_agg(max,
(j.date_finished
for j in all_phases if j.date_finished))
else:
job.date_finished = None
if job.date_started and job.date_finished:
job.duration = int(
(job.date_finished - job.date_started).total_seconds() * 1000)
else:
job.duration = None
# if any phases are marked as failing, fail the build
if any(j.result is Result.failed for j in all_phases):
job.result = Result.failed
# If any test cases were marked as failing, fail the build.
# The exception is if the only failing test case occurred in a JobStep that
# had an infra failure. In this case we can't trust the test case result as
# being meaningful and so we ignore these.
elif TestCase.query.join(JobStep, JobStep.id == TestCase.step_id).filter(
TestCase.result == Result.failed, TestCase.job_id == job.id,
JobStep.result != Result.infra_failed).first():
job.result = Result.failed
# if we've finished all phases, use the best result available
elif is_finished:
# Sets the final job result.
implementation.validate(job=job)
else:
job.result = Result.unknown
if is_finished:
job.status = Status.finished
else:
# ensure we dont set the status to finished unless it actually is
new_status = aggregate_status((j.status for j in all_phases))
if new_status != Status.finished:
job.status = new_status
elif job.status == Status.finished:
job.status = Status.in_progress
current_app.logger.exception(
'Job incorrectly marked as finished: %s', job.id)
if db.session.is_modified(job):
job.date_modified = datetime.utcnow()
db.session.add(job)
db.session.commit()
if not is_finished:
raise sync_job.NotFinished
try:
aggregate_job_stat(job, 'test_count')
aggregate_job_stat(job, 'test_duration')
aggregate_job_stat(job, 'test_failures')
aggregate_job_stat(job, 'test_rerun_count')
aggregate_job_stat(job, 'tests_missing')
aggregate_job_stat(job, 'lines_covered')
aggregate_job_stat(job, 'lines_uncovered')
aggregate_job_stat(job, 'diff_lines_covered')
aggregate_job_stat(job, 'diff_lines_uncovered')
except Exception:
current_app.logger.exception(
'Failing recording aggregate stats for job %s', job.id)
fire_signal.delay(
signal='job.finished',
kwargs={'job_id': job.id.hex},
)
if jobplan:
queue.delay('update_project_plan_stats',
kwargs={
'project_id': job.project_id.hex,
'plan_id': jobplan.plan_id.hex,
},
countdown=1)
def sync_build(build_id):
"""
Synchronizing the build happens continuously until all jobs have reported in
as finished or have failed/aborted.
This task is responsible for:
- Checking in with jobs
- Aborting/retrying them if they're beyond limits
- Aggregating the results from jobs into the build itself
"""
build = Build.query.get(build_id)
if not build:
return
if build.status == Status.finished:
return
all_jobs = list(Job.query.filter(
Job.build_id == build_id,
))
is_finished = sync_build.verify_all_children() == Status.finished
if any(p.status != Status.finished for p in all_jobs):
is_finished = False
build.date_started = safe_agg(
min, (j.date_started for j in all_jobs if j.date_started))
if is_finished:
build.date_finished = safe_agg(
max, (j.date_finished for j in all_jobs if j.date_finished))
else:
build.date_finished = None
if build.date_started and build.date_finished:
build.duration = int((build.date_finished - build.date_started).total_seconds() * 1000)
else:
build.duration = None
if any(j.result is Result.failed for j in all_jobs):
build.result = Result.failed
elif is_finished:
build.result = aggregate_result((j.result for j in all_jobs))
else:
build.result = Result.unknown
if is_finished:
build.status = Status.finished
else:
# ensure we dont set the status to finished unless it actually is
new_status = aggregate_status((j.status for j in all_jobs))
if new_status != Status.finished:
build.status = new_status
if db.session.is_modified(build):
build.date_modified = datetime.utcnow()
db.session.add(build)
db.session.commit()
if not is_finished:
raise sync_build.NotFinished
try:
aggregate_build_stat(build, 'test_count')
aggregate_build_stat(build, 'test_duration')
aggregate_build_stat(build, 'test_failures')
aggregate_build_stat(build, 'test_rerun_count')
aggregate_build_stat(build, 'tests_missing')
aggregate_build_stat(build, 'lines_covered')
aggregate_build_stat(build, 'lines_uncovered')
aggregate_build_stat(build, 'diff_lines_covered')
aggregate_build_stat(build, 'diff_lines_uncovered')
except Exception:
current_app.logger.exception('Failing recording aggregate stats for build %s', build.id)
fire_signal.delay(
signal='build.finished',
kwargs={'build_id': build.id.hex},
)
queue.delay('update_project_stats', kwargs={
'project_id': build.project_id.hex,
}, countdown=1)
def sync_job(job_id):
job = Job.query.get(job_id)
if not job:
return
if job.status == Status.finished:
return
# TODO(dcramer): we make an assumption that there is a single step
job_plan = JobPlan.query.options(
subqueryload_all('plan.steps')
).filter(
JobPlan.job_id == job.id,
).join(Plan).first()
try:
if not job_plan:
raise UnrecoverableException('Got sync_job task without job plan: %s' % (job.id,))
try:
step = job_plan.plan.steps[0]
except IndexError:
raise UnrecoverableException('Missing steps for plan')
implementation = step.get_implementation()
if has_timed_out(job, job_plan):
remaining_steps = list(JobStep.query.filter(
JobStep.status != Status.finished,
JobStep.job_id == job.id,
))
implementation.cancel(job=job)
for step in remaining_steps:
step.result = Result.failed
step.status = Status.finished
db.session.add(step)
try_create(FailureReason, {
'step_id': step.id,
'job_id': job.id,
'build_id': job.build_id,
'project_id': job.project_id,
'reason': 'timeout'
})
# ensure the job result actually reflects a failure
job.result = Result.failed
job.status = Status.finished
db.session.add(job)
else:
implementation.update(job=job)
except UnrecoverableException:
job.status = Status.finished
job.result = Result.aborted
current_app.logger.exception('Unrecoverable exception syncing %s', job.id)
is_finished = sync_job.verify_all_children() == Status.finished
if is_finished:
job.status = Status.finished
db.session.flush()
all_phases = list(job.phases)
# propagate changes to any phases as they live outside of the
# normalize synchronization routines
sync_job_phases(job, all_phases)
job.date_started = safe_agg(
min, (j.date_started for j in all_phases if j.date_started))
if is_finished:
job.date_finished = safe_agg(
max, (j.date_finished for j in all_phases if j.date_finished))
else:
job.date_finished = None
if job.date_started and job.date_finished:
job.duration = int((job.date_finished - job.date_started).total_seconds() * 1000)
else:
job.duration = None
# if any phases are marked as failing, fail the build
if any(j.result is Result.failed for j in all_phases):
job.result = Result.failed
# if any test cases were marked as failing, fail the build
elif TestCase.query.filter(TestCase.result == Result.failed, TestCase.job_id == job.id).first():
job.result = Result.failed
# if we've finished all phases, use the best result available
elif is_finished:
job.result = safe_agg(max, (j.result for j in all_phases))
else:
job.result = Result.unknown
if is_finished:
job.status = Status.finished
elif any(j.status is not Status.queued for j in all_phases):
job.status = Status.in_progress
else:
job.status = Status.queued
if db.session.is_modified(job):
job.date_modified = datetime.utcnow()
db.session.add(job)
db.session.commit()
if not is_finished:
raise sync_job.NotFinished
try:
aggregate_job_stat(job, 'test_count')
aggregate_job_stat(job, 'test_duration')
aggregate_job_stat(job, 'test_failures')
aggregate_job_stat(job, 'test_rerun_count')
aggregate_job_stat(job, 'tests_missing')
aggregate_job_stat(job, 'lines_covered')
aggregate_job_stat(job, 'lines_uncovered')
aggregate_job_stat(job, 'diff_lines_covered')
aggregate_job_stat(job, 'diff_lines_uncovered')
except Exception:
current_app.logger.exception('Failing recording aggregate stats for job %s', job.id)
fire_signal.delay(
signal='job.finished',
kwargs={'job_id': job.id.hex},
)
if job_plan:
queue.delay('update_project_plan_stats', kwargs={
'project_id': job.project_id.hex,
'plan_id': job_plan.plan_id.hex,
}, countdown=1)
def sync_job(job_id):
"""
Updates jobphase and job statuses based on the status of the constituent jobsteps.
"""
job = Job.query.get(job_id)
if not job:
return
if job.status == Status.finished:
return
jobplan, implementation = JobPlan.get_build_step_for_job(job_id=job.id)
try:
implementation.update(job=job)
except UnrecoverableException:
job.status = Status.finished
job.result = Result.infra_failed
current_app.logger.exception('Unrecoverable exception syncing %s', job.id)
all_phases = list(job.phases)
# propagate changes to any phases as they live outside of the
# normalize synchronization routines
sync_job_phases(job, all_phases, implementation)
is_finished = sync_job.verify_all_children() == Status.finished
if any(p.status != Status.finished for p in all_phases):
is_finished = False
job.date_started = safe_agg(
min, (j.date_started for j in all_phases if j.date_started))
if is_finished:
job.date_finished = safe_agg(
max, (j.date_finished for j in all_phases if j.date_finished))
else:
job.date_finished = None
if job.date_started and job.date_finished:
job.duration = int((job.date_finished - job.date_started).total_seconds() * 1000)
else:
job.duration = None
# if any phases are marked as failing, fail the build
if any(j.result is Result.failed for j in all_phases):
job.result = Result.failed
# If any test cases were marked as failing, fail the build.
# The exception is if the only failing test case occurred in a JobStep that
# had an infra failure. In this case we can't trust the test case result as
# being meaningful and so we ignore these.
elif TestCase.query.join(JobStep, JobStep.id == TestCase.step_id).filter(
TestCase.result == Result.failed, TestCase.job_id == job.id,
JobStep.result != Result.infra_failed
).first():
job.result = Result.failed
# if we've finished all phases, use the best result available
elif is_finished:
# Sets the final job result.
implementation.validate(job=job)
else:
job.result = Result.unknown
if is_finished:
job.status = Status.finished
else:
# ensure we dont set the status to finished unless it actually is
new_status = aggregate_status((j.status for j in all_phases))
if new_status != Status.finished:
job.status = new_status
elif job.status == Status.finished:
job.status = Status.in_progress
current_app.logger.exception('Job incorrectly marked as finished: %s', job.id)
if db.session.is_modified(job):
job.date_modified = datetime.utcnow()
db.session.add(job)
db.session.commit()
if not is_finished:
raise sync_job.NotFinished
try:
aggregate_job_stat(job, 'test_count')
aggregate_job_stat(job, 'test_duration')
aggregate_job_stat(job, 'test_failures')
aggregate_job_stat(job, 'test_rerun_count')
aggregate_job_stat(job, 'tests_missing')
aggregate_job_stat(job, 'lines_covered')
aggregate_job_stat(job, 'lines_uncovered')
aggregate_job_stat(job, 'diff_lines_covered')
aggregate_job_stat(job, 'diff_lines_uncovered')
except Exception:
current_app.logger.exception('Failing recording aggregate stats for job %s', job.id)
fire_signal.delay(
signal='job.finished',
kwargs={'job_id': job.id.hex},
)
if jobplan:
queue.delay('update_project_plan_stats', kwargs={
'project_id': job.project_id.hex,
'plan_id': jobplan.plan_id.hex,
}, countdown=1)
def sync_job(job_id):
job = Job.query.get(job_id)
if not job:
return
if job.status == Status.finished:
return
# TODO(dcramer): we make an assumption that there is a single step
jobplan, implementation = JobPlan.get_build_step_for_job(job_id=job.id)
try:
implementation.update(job=job)
except UnrecoverableException:
job.status = Status.finished
job.result = Result.aborted
current_app.logger.exception('Unrecoverable exception syncing %s',
job.id)
is_finished = sync_job.verify_all_children() == Status.finished
if is_finished:
job.status = Status.finished
db.session.flush()
all_phases = list(job.phases)
# propagate changes to any phases as they live outside of the
# normalize synchronization routines
sync_job_phases(job, all_phases)
job.date_started = safe_agg(min, (j.date_started
for j in all_phases if j.date_started))
if is_finished:
job.date_finished = safe_agg(max,
(j.date_finished
for j in all_phases if j.date_finished))
else:
job.date_finished = None
if job.date_started and job.date_finished:
job.duration = int(
(job.date_finished - job.date_started).total_seconds() * 1000)
else:
job.duration = None
# if any phases are marked as failing, fail the build
if any(j.result is Result.failed for j in all_phases):
job.result = Result.failed
# if any test cases were marked as failing, fail the build
elif TestCase.query.filter(TestCase.result == Result.failed,
TestCase.job_id == job.id).first():
job.result = Result.failed
# if we've finished all phases, use the best result available
elif is_finished:
job.result = safe_agg(max, (j.result for j in all_phases))
else:
job.result = Result.unknown
if is_finished:
job.status = Status.finished
elif any(j.status is not Status.queued for j in all_phases):
job.status = Status.in_progress
else:
job.status = Status.queued
if db.session.is_modified(job):
job.date_modified = datetime.utcnow()
db.session.add(job)
db.session.commit()
if not is_finished:
raise sync_job.NotFinished
try:
aggregate_job_stat(job, 'test_count')
aggregate_job_stat(job, 'test_duration')
aggregate_job_stat(job, 'test_failures')
aggregate_job_stat(job, 'test_rerun_count')
aggregate_job_stat(job, 'tests_missing')
aggregate_job_stat(job, 'lines_covered')
aggregate_job_stat(job, 'lines_uncovered')
aggregate_job_stat(job, 'diff_lines_covered')
aggregate_job_stat(job, 'diff_lines_uncovered')
except Exception:
current_app.logger.exception(
'Failing recording aggregate stats for job %s', job.id)
fire_signal.delay(
signal='job.finished',
kwargs={'job_id': job.id.hex},
)
if jobplan:
queue.delay('update_project_plan_stats',
kwargs={
'project_id': job.project_id.hex,
'plan_id': jobplan.plan_id.hex,
},
countdown=1)
def sync_build(build_id):
"""
Synchronizing the build happens continuously until all jobs have reported in
as finished or have failed/aborted.
This task is responsible for:
- Checking in with jobs
- Aborting/retrying them if they're beyond limits
- Aggregating the results from jobs into the build itself
"""
build = Build.query.get(build_id)
if not build:
return
if build.status == Status.finished:
return
all_jobs = list(Job.query.filter(Job.build_id == build_id, ))
is_finished = sync_build.verify_all_children() == Status.finished
build.date_started = safe_agg(min, (j.date_started
for j in all_jobs if j.date_started))
if is_finished:
build.date_finished = safe_agg(max,
(j.date_finished
for j in all_jobs if j.date_finished))
else:
build.date_finished = None
if build.date_started and build.date_finished:
build.duration = int(
(build.date_finished - build.date_started).total_seconds() * 1000)
else:
build.duration = None
if any(j.result is Result.failed for j in all_jobs):
build.result = Result.failed
elif is_finished:
build.result = safe_agg(max, (j.result for j in all_jobs))
else:
build.result = Result.unknown
if is_finished:
build.status = Status.finished
elif any(j.status is not Status.queued for j in all_jobs):
build.status = Status.in_progress
else:
build.status = Status.queued
if db.session.is_modified(build):
build.date_modified = datetime.utcnow()
db.session.add(build)
db.session.commit()
if not is_finished:
raise sync_build.NotFinished
try:
aggregate_build_stat(build, 'test_count')
aggregate_build_stat(build, 'test_duration')
aggregate_build_stat(build, 'test_failures')
aggregate_build_stat(build, 'test_rerun_count')
aggregate_build_stat(build, 'tests_missing')
aggregate_build_stat(build, 'lines_covered')
aggregate_build_stat(build, 'lines_uncovered')
aggregate_build_stat(build, 'diff_lines_covered')
aggregate_build_stat(build, 'diff_lines_uncovered')
except Exception:
current_app.logger.exception(
'Failing recording aggregate stats for build %s', build.id)
fire_signal.delay(
signal='build.finished',
kwargs={'build_id': build.id.hex},
)
queue.delay('update_project_stats',
kwargs={
'project_id': build.project_id.hex,
},
countdown=1)
def sync_job(job_id):
job = Job.query.get(job_id)
if not job:
return
if job.status == Status.finished:
return
# TODO(dcramer): we make an assumption that there is a single step
job_plan = JobPlan.query.options(
subqueryload_all('plan.steps')
).filter(
JobPlan.job_id == job.id,
).join(Plan).first()
try:
if not job_plan:
raise UnrecoverableException('Got sync_job task without job plan: %s' % (job.id,))
try:
step = job_plan.plan.steps[0]
except IndexError:
raise UnrecoverableException('Missing steps for plan')
implementation = step.get_implementation()
implementation.update(job=job)
except UnrecoverableException:
job.status = Status.finished
job.result = Result.aborted
current_app.logger.exception('Unrecoverable exception syncing %s', job.id)
is_finished = sync_job.verify_all_children() == Status.finished
if is_finished:
job.status = Status.finished
all_phases = list(job.phases)
job.date_started = safe_agg(
min, (j.date_started for j in all_phases if j.date_started))
if is_finished:
job.date_finished = safe_agg(
max, (j.date_finished for j in all_phases if j.date_finished))
else:
job.date_finished = None
if job.date_started and job.date_finished:
job.duration = int((job.date_finished - job.date_started).total_seconds() * 1000)
else:
job.duration = None
# if any phases are marked as failing, fail the build
if any(j.result is Result.failed for j in all_phases):
job.result = Result.failed
# if any test cases were marked as failing, fail the build
elif TestCase.query.filter(TestCase.result == Result.failed, TestCase.job_id == job.id).first():
job.result = Result.failed
# if we've finished all phases, use the best result available
elif is_finished:
job.result = safe_agg(
max, (j.result for j in all_phases), Result.unknown)
else:
job.result = Result.unknown
if is_finished:
job.status = Status.finished
elif any(j.status is Status.in_progress for j in all_phases):
job.status = Status.in_progress
else:
job.status = Status.queued
if db.session.is_modified(job):
job.date_modified = datetime.utcnow()
db.session.add(job)
db.session.commit()
publish_job_update(job)
if not is_finished:
raise sync_job.NotFinished
_record_tests_missing(job)
queue.delay('notify_job_finished', kwargs={
'job_id': job.id.hex,
})
if job_plan:
queue.delay('update_project_plan_stats', kwargs={
'project_id': job.project_id.hex,
'plan_id': job_plan.plan_id.hex,
}, countdown=1)
def _sync_step_from_active(self, step):
try:
job_name = step.data['job_name']
build_no = step.data['build_no']
except KeyError:
raise UnrecoverableException('Missing Jenkins job information')
try:
item = self._get_response('/job/{}/{}'.format(
job_name, build_no))
except NotFound:
raise UnrecoverableException('Unable to find job in Jenkins')
# TODO(dcramer): we're doing a lot of work here when we might
# not need to due to it being sync'd previously
node, _ = get_or_create(Node, where={
'label': item['builtOn'],
})
step.node = node
step.label = item['fullDisplayName']
step.date_started = datetime.utcfromtimestamp(
item['timestamp'] / 1000)
if item['building']:
step.status = Status.in_progress
else:
step.status = Status.finished
step.result = RESULT_MAP[item['result']]
# values['duration'] = item['duration'] or None
step.date_finished = datetime.utcfromtimestamp(
(item['timestamp'] + item['duration']) / 1000)
# step.data.update({
# 'backend': {
# 'uri': item['url'],
# 'label': item['fullDisplayName'],
# }
# })
db.session.add(step)
db.session.commit()
# TODO(dcramer): we shoudl abstract this into a sync_phase
phase = step.phase
if not phase.date_started:
phase.date_started = safe_agg(
min, (s.date_started for s in phase.steps), step.date_started)
db.session.add(phase)
if phase.status != step.status:
phase.status = step.status
db.session.add(phase)
if step.status == Status.finished:
phase.status = Status.finished
phase.date_finished = safe_agg(
max, (s.date_finished for s in phase.steps), step.date_finished)
if any(s.result is Result.failed for s in phase.steps):
phase.result = Result.failed
else:
phase.result = safe_agg(
max, (s.result for s in phase.steps), Result.unknown)
db.session.add(phase)
db.session.commit()
if step.status != Status.finished:
return
# sync artifacts
for artifact in item.get('artifacts', ()):
artifact, created = get_or_create(Artifact, where={
'step': step,
'name': artifact['fileName'],
}, defaults={
'project': step.project,
'job': step.job,
'data': artifact,
})
db.session.commit()
sync_artifact.delay_if_needed(
artifact_id=artifact.id.hex,
task_id=artifact.id.hex,
parent_task_id=step.id.hex,
)
# sync test results
try:
with db.session.begin_nested():
self._sync_test_results(
step=step,
job_name=job_name,
build_no=build_no,
)
except Exception:
self.logger.exception(
'Failed to sync test results for %s #%s', job_name, build_no)
else:
db.session.commit()
# sync console log
try:
result = True
while result:
result = self._sync_log(
jobstep=step,
name=step.label,
job_name=job_name,
build_no=build_no,
)
except Exception:
db.session.rollback()
current_app.logger.exception(
'Unable to sync console log for job step %r',
step.id.hex)
