More exception handling. Remove unused database models. New common view structure!

This commit is contained in:
Patrick Jentsch
2020-11-13 10:01:51 +01:00
parent cb9da5c7dd
commit 5a06a6b241
45 changed files with 692 additions and 1005 deletions

View File

@ -11,15 +11,11 @@ def check_corpora():
corpora = Corpus.query.all()
for corpus in filter(lambda corpus: corpus.status == 'submitted', corpora):
corpus_utils.create_build_corpus_service(corpus)
for corpus in filter(lambda corpus: (corpus.status == 'queued'
or corpus.status == 'running'),
corpora):
for corpus in filter(lambda corpus: corpus.status in ['queued', 'running'], corpora): # noqa
corpus_utils.checkout_build_corpus_service(corpus)
for corpus in filter(lambda corpus: corpus.status == 'start analysis',
corpora):
for corpus in filter(lambda corpus: corpus.status == 'start analysis', corpora): # noqa
corpus_utils.create_cqpserver_container(corpus)
for corpus in filter(lambda corpus: corpus.status == 'stop analysis',
corpora):
for corpus in filter(lambda corpus: corpus.status == 'stop analysis', corpora): # noqa
corpus_utils.remove_cqpserver_container(corpus)
db.session.commit()
@ -28,8 +24,6 @@ def check_jobs():
jobs = Job.query.all()
for job in filter(lambda job: job.status == 'submitted', jobs):
job_utils.create_job_service(job)
for job in filter(lambda job: job.status == 'queued', jobs):
job_utils.checkout_job_service(job)
for job in filter(lambda job: job.status == 'running', jobs):
for job in filter(lambda job: job.status in ['queued', 'running'], jobs):
job_utils.checkout_job_service(job)
db.session.commit()

View File

@ -1,4 +1,3 @@
from flask import current_app
from . import docker_client
import docker
import logging
@ -7,20 +6,14 @@ import shutil
def create_build_corpus_service(corpus):
corpus_dir = os.path.join(current_app.config['DATA_DIR'],
str(corpus.user_id),
'corpora',
str(corpus.id))
corpus_data_dir = os.path.join(corpus_dir, 'data')
corpus_file = os.path.join(corpus_dir, 'merged', 'corpus.vrt')
corpus_registry_dir = os.path.join(corpus_dir, 'registry')
if os.path.exists(corpus_data_dir):
shutil.rmtree(corpus_data_dir)
if os.path.exists(corpus_registry_dir):
shutil.rmtree(corpus_registry_dir)
corpus_data_dir = os.path.join(corpus.path, 'data')
shutil.rmtree(corpus_data_dir, ignore_errors=True)
os.mkdir(corpus_data_dir)
corpus_registry_dir = os.path.join(corpus.path, 'registry')
shutil.rmtree(corpus_registry_dir, ignore_errors=True)
os.mkdir(corpus_registry_dir)
service_args = {
corpus_file = os.path.join(corpus.path, 'merged', 'corpus.vrt')
service_kwargs = {
'command': 'docker-entrypoint.sh build-corpus',
'constraints': ['node.role==worker'],
'labels': {'origin': 'nopaque',
@ -32,30 +25,34 @@ def create_build_corpus_service(corpus):
'name': 'build-corpus_{}'.format(corpus.id),
'restart_policy': docker.types.RestartPolicy()
}
service_image = \
'gitlab.ub.uni-bielefeld.de:4567/sfb1288inf/cqpserver:latest'
service_image = 'gitlab.ub.uni-bielefeld.de:4567/sfb1288inf/cqpserver:latest' # noqa
try:
docker_client.services.create(service_image, **service_args)
docker_client.services.create(service_image, **service_kwargs)
except docker.errors.APIError as e:
logging.error('create_build_corpus_service({}): '.format(corpus.id)
+ '{} (status: {} -> failed)'.format(e, corpus.status))
corpus.status = 'failed'
logging.error('Create "{}" service raised '.format(service_kwargs['name']) # noqa
+ '[docker-APIError] The server returned an error. '
+ 'Details: {}'.format(e))
else:
corpus.status = 'queued'
finally:
# TODO: send email
pass
def checkout_build_corpus_service(corpus):
service_name = 'build-corpus_{}'.format(corpus.id)
try:
service = docker_client.services.get(service_name)
except docker.errors.NotFound as e:
logging.error('checkout_build_corpus_service({}):'.format(corpus.id)
+ ' {} (stauts: {} -> failed)'.format(e, corpus.status))
except docker.errors.NotFound:
logging.error('Get "{}" service raised '.format(service_name)
+ '[docker-NotFound] The service does not exist. '
+ '(corpus.status: {} -> failed)'.format(corpus.status))
corpus.status = 'failed'
# TODO: handle docker.errors.APIError and docker.errors.InvalidVersion
except docker.errors.APIError as e:
logging.error('Get "{}" service raised '.format(service_name)
+ '[docker-APIError] The server returned an error. '
+ 'Details: {}'.format(e))
except docker.errors.InvalidVersion:
logging.error('Get "{}" service raised '.format(service_name)
+ '[docker-InvalidVersion] One of the arguments is '
+ 'not supported with the current API version.')
else:
service_tasks = service.tasks()
if not service_tasks:
@ -63,25 +60,23 @@ def checkout_build_corpus_service(corpus):
task_state = service_tasks[0].get('Status').get('State')
if corpus.status == 'queued' and task_state != 'pending':
corpus.status = 'running'
elif corpus.status == 'running' and task_state == 'complete':
service.remove()
corpus.status = 'prepared'
elif corpus.status == 'running' and task_state == 'failed':
service.remove()
corpus.status = task_state
finally:
# TODO: send email
pass
elif corpus.status == 'running' and task_state in ['complete', 'failed']: # noqa
try:
service.remove()
except docker.errors.APIError as e:
logging.error('Remove "{}" service raised '.format(service_name) # noqa
+ '[docker-APIError] The server returned an error. ' # noqa
+ 'Details: {}'.format(e))
return
else:
corpus.status = 'prepared' if task_state == 'complete' \
else 'failed'
def create_cqpserver_container(corpus):
corpus_dir = os.path.join(current_app.config['DATA_DIR'],
str(corpus.user_id),
'corpora',
str(corpus.id))
corpus_data_dir = os.path.join(corpus_dir, 'data')
corpus_registry_dir = os.path.join(corpus_dir, 'registry')
container_args = {
corpus_data_dir = os.path.join(corpus.path, 'data')
corpus_registry_dir = os.path.join(corpus.path, 'registry')
container_kwargs = {
'command': 'cqpserver',
'detach': True,
'volumes': [corpus_data_dir + ':/corpora/data:rw',
@ -89,20 +84,43 @@ def create_cqpserver_container(corpus):
'name': 'cqpserver_{}'.format(corpus.id),
'network': 'nopaque_default'
}
container_image = \
'gitlab.ub.uni-bielefeld.de:4567/sfb1288inf/cqpserver:latest'
container_image = 'gitlab.ub.uni-bielefeld.de:4567/sfb1288inf/cqpserver:latest' # noqa
# Check if a cqpserver container already exists. If this is the case,
# remove it and create a new one
try:
container = docker_client.containers.get(container_args['name'])
container = docker_client.containers.get(container_kwargs['name'])
except docker.errors.NotFound:
pass
except docker.errors.DockerException:
except docker.errors.APIError as e:
logging.error('Get "{}" container raised '.format(container_kwargs['name'])
+ '[docker-APIError] The server returned an error. '
+ 'Details: {}'.format(e))
return
else:
container.remove(force=True)
try:
container.remove(force=True)
except docker.errors.APIError as e:
logging.error('Remove "{}" container raised '.format(container_kwargs['name'])
+ '[docker-APIError] The server returned an error. '
+ 'Details: {}'.format(e))
return
try:
docker_client.containers.run(container_image, **container_args)
except docker.errors.DockerException:
return
docker_client.containers.run(container_image, **container_kwargs)
except docker.errors.ContainerError:
# This case should not occur, because detach is True.
logging.error('Run "{}" container raised '.format(container_kwargs['name'])
+ '[docker-ContainerError] The container exits with a '
+ 'non-zero exit code and detach is False.')
corpus.status = 'failed'
except docker.errors.ImageNotFound:
logging.error('Run "{}" container raised '.format(container_kwargs['name'])
+ '[docker-ImageNotFound] The specified image does not '
+ 'exist.')
corpus.status = 'failed'
except docker.errors.APIError as e:
logging.error('Run "{}" container raised '.format(container_kwargs['name'])
+ '[docker-APIError] The server returned an error. '
+ 'Details: {}'.format(e))
else:
corpus.status = 'analysing'
@ -113,8 +131,17 @@ def remove_cqpserver_container(corpus):
container = docker_client.containers.get(container_name)
except docker.errors.NotFound:
pass
except docker.errors.DockerException:
except docker.errors.APIError as e:
logging.error('Get "{}" container raised '.format(container_name)
+ '[docker-APIError] The server returned an error. '
+ 'Details: {}'.format(e))
return
else:
container.remove(force=True)
try:
container.remove(force=True)
except docker.errors.APIError as e:
logging.error('Remove "{}" container raised '.format(container_name)
+ '[docker-APIError] The server returned an error. '
+ 'Details: {}'.format(e))
return
corpus.status = 'prepared'

View File

@ -1,7 +1,7 @@
from datetime import datetime
from flask import current_app
from . import docker_client
from .. import db
from ..email import create_message, send
from ..models import JobResult
import docker
import logging
@ -10,51 +10,60 @@ import os
def create_job_service(job):
job_dir = os.path.join(current_app.config['DATA_DIR'],
str(job.user_id),
'jobs',
str(job.id))
cmd = '{} -i /files -o /files/output'.format(job.service)
if job.service == 'file-setup':
cmd += ' -f {}'.format(job.secure_filename)
cmd += ' --log-dir /files'
cmd += ' --zip [{}]_{}'.format(job.service, job.secure_filename)
cmd += ' ' + ' '.join(json.loads(job.service_args))
service_args = {'command': cmd,
'constraints': ['node.role==worker'],
'labels': {'origin': 'nopaque',
'type': 'service.{}'.format(job.service),
'job_id': str(job.id)},
'mounts': [job_dir + ':/files:rw'],
'name': 'job_{}'.format(job.id),
'resources': docker.types.Resources(
cpu_reservation=job.n_cores * (10 ** 9),
mem_reservation=job.mem_mb * (10 ** 6)),
'restart_policy': docker.types.RestartPolicy()}
service_kwargs = {'command': cmd,
'constraints': ['node.role==worker'],
'labels': {'origin': 'nopaque',
'type': 'service.{}'.format(job.service),
'job_id': str(job.id)},
'mounts': [job.path + ':/files:rw'],
'name': 'job_{}'.format(job.id),
'resources': docker.types.Resources(
cpu_reservation=job.n_cores * (10 ** 9),
mem_reservation=job.mem_mb * (10 ** 6)
),
'restart_policy': docker.types.RestartPolicy()}
service_image = ('gitlab.ub.uni-bielefeld.de:4567/sfb1288inf/'
+ job.service + ':' + job.service_version)
try:
docker_client.services.create(service_image, **service_args)
docker_client.services.create(service_image, **service_kwargs)
except docker.errors.APIError as e:
logging.error('create_job_service({}): {} '.format(job.id, e)
+ '(status: {} -> failed)'.format(job.status))
job.status = 'failed'
logging.error('Create "{}" service raised '.format(service_kwargs['name']) # noqa
+ '[docker-APIError] The server returned an error. '
+ 'Details: {}'.format(e))
else:
job.status = 'queued'
finally:
# TODO: send email
pass
msg = create_message(
job.creator.email,
'Status update for your Job "{}"'.format(job.title),
'tasks/email/notification',
job=job
)
send(msg)
def checkout_job_service(job):
service_name = 'job_{}'.format(job.id)
try:
service = docker_client.services.get(service_name)
except docker.errors.NotFound as e:
logging.error('checkout_job_service({}): {} '.format(job.id, e)
+ '(status: {} -> submitted)'.format(job.status))
job.status = 'submitted'
# TODO: handle docker.errors.APIError and docker.errors.InvalidVersion
except docker.errors.NotFound:
logging.error('Get "{}" service raised '.format(service_name)
+ '[docker-NotFound] The service does not exist. '
+ '(job.status: {} -> failed)'.format(job.status))
job.status = 'failed'
except docker.errors.APIError as e:
logging.error('Get "{}" service raised '.format(service_name)
+ '[docker-APIError] The server returned an error. '
+ 'Details: {}'.format(e))
except docker.errors.InvalidVersion:
logging.error('Get "{}" service raised '.format(service_name)
+ '[docker-InvalidVersion] One of the arguments is '
+ 'not supported with the current API version.')
else:
service_tasks = service.tasks()
if not service_tasks:
@ -62,22 +71,16 @@ def checkout_job_service(job):
task_state = service_tasks[0].get('Status').get('State')
if job.status == 'queued' and task_state != 'pending':
job.status = 'running'
elif job.status == 'queued' and task_state == 'complete':
elif job.status == 'running' and task_state == 'complete':
service.remove()
job.end_date = datetime.utcnow()
job.status = task_state
if task_state == 'complete':
results_dir = os.path.join(current_app.config['DATA_DIR'],
str(job.user_id),
'jobs',
str(job.id),
'output')
results = filter(lambda x: x.endswith('.zip'),
os.listdir(results_dir))
for result in results:
job_result = JobResult(dir=results_dir,
filename=result,
job_id=job.id)
job_results_dir = os.path.join(job.path, 'output')
job_results = filter(lambda x: x.endswith('.zip'),
os.listdir(job_results_dir))
for job_result in job_results:
job_result = JobResult(filename=job_result, job=job)
db.session.add(job_result)
elif job.status == 'running' and task_state == 'failed':
service.remove()
@ -85,6 +88,13 @@ def checkout_job_service(job):
job.status = task_state
finally:
# TODO: send email
msg = create_message(
job.creator.email,
'[nopaque] Status update for your Job "{}"'.format(job.title),
'tasks/email/notification',
job=job
)
send(msg)
pass