def job_not_assigned(request, jobid, path):
if request.GET.get('format') == 'json':
result = {'status': -1, 'message': ''}
try:
get_api(request.user, request.jt).get_job(jobid=jobid)
result['status'] = 0
except ApplicationNotRunning, e:
result['status'] = 1
except Exception, e:
result['message'] = _('Error polling job %s: %s') % (jobid, e)
def job_not_assigned(request, jobid, path):
if request.GET.get("format") == "json":
result = {"status": -1, "message": ""}
try:
get_api(request.user, request.jt).get_job(jobid=jobid)
result["status"] = 0
except ApplicationNotRunning, e:
result["status"] = 1
except Exception, e:
result["message"] = _("Error polling job %s: %s") % (jobid, e)
def single_tracker(request, trackerid):
jt = get_api(request.user, request.jt)
try:
tracker = jt.get_tracker(trackerid)
except Exception, e:
raise PopupException(_('The tracker could not be contacted.'), detail=e)
def jobs(request):
user = request.POST.get('user', request.user.username)
state = request.POST.get('state')
text = request.POST.get('text')
retired = request.POST.get('retired')
time_value = request.POST.get('time_value', 7)
time_unit = request.POST.get('time_unit', 'days')
if request.POST.get('format') == 'json':
try:
# Limit number of jobs to be 1000
jobs = get_api(request.user, request.jt).get_jobs(user=request.user, username=user, state=state, text=text,
retired=retired, limit=1000, time_value=int(time_value), time_unit=time_unit)
except Exception, ex:
ex_message = str(ex)
if 'Connection refused' in ex_message or 'standby RM' in ex_message:
raise PopupException(_('Resource Manager cannot be contacted or might be down.'))
elif 'Could not connect to' in ex_message:
raise PopupException(_('Job Tracker cannot be contacted or might be down.'))
else:
raise PopupException(ex)
json_jobs = {
'jobs': [massage_job_for_json(job, request) for job in jobs],
}
return JsonResponse(json_jobs, encoder=JSONEncoderForHTML)
def job_attempt_logs_json(request, job, attempt_index=0, name='syslog', offset=LOG_OFFSET_BYTES, is_embeddable=False):
"""For async log retrieval as Yarn servers are very slow"""
log_link = None
response = {'status': -1}
try:
jt = get_api(request.user, request.jt)
app = jt.get_application(job.jobId)
if app['applicationType'] == 'MAPREDUCE':
if app['finalStatus'] in ('SUCCEEDED', 'FAILED', 'KILLED'):
attempt_index = int(attempt_index)
if not job.job_attempts['jobAttempt']:
response = {'status': 0, 'log': _('Job has no tasks')}
else:
attempt = job.job_attempts['jobAttempt'][attempt_index]
log_link = attempt['logsLink']
# Reformat log link to use YARN RM, replace node addr with node ID addr
log_link = log_link.replace(attempt['nodeHttpAddress'], attempt['nodeId'])
elif app['state'] == 'RUNNING':
log_link = app['amContainerLogs']
elif app.get('amContainerLogs'):
log_link = app.get('amContainerLogs')
except (KeyError, RestException), e:
raise KeyError(_("Cannot find job attempt '%(id)s'.") % {'id': job.jobId}, e)
def jobs(request):
user = request.POST.get('user', request.user.username)
state = request.POST.get('state')
text = request.POST.get('text')
retired = request.POST.get('retired')
if request.POST.get('format') == 'json':
try:
# Limit number of jobs to be 1000
jobs = get_api(request.user,
request.jt).get_jobs(user=request.user,
username=user,
state=state,
text=text,
retired=retired,
limit=1000)
except Exception, ex:
ex_message = str(ex)
if 'Connection refused' in ex_message or 'standby RM' in ex_message:
raise PopupException(
_('Resource Manager cannot be contacted or might be down.')
)
elif 'Could not connect to' in ex_message:
raise PopupException(
_('Job Tracker cannot be contacted or might be down.'))
else:
raise PopupException(ex)
json_jobs = {
'jobs': [massage_job_for_json(job, request) for job in jobs],
}
return JsonResponse(json_jobs, encoder=JSONEncoderForHTML)
def single_task(request, job, taskid):
jt = get_api(request.user, request.jt)
job_link = jt.get_job_link(job.jobId)
task = job_link.get_task(taskid)
return render("task.mako", request, {'task': task, 'joblnk': job_link})
def single_task(request, job, taskid):
jt = get_api(request.user, request.jt)
job_link = jt.get_job_link(job.jobId)
task = job_link.get_task(taskid)
return render("task.mako", request, {"task": task, "joblnk": job_link})
def decorate(request, *args, **kwargs):
jobid = kwargs['job']
try:
job = get_api(request.user, request.jt).get_job(jobid=jobid)
except ApplicationNotRunning, e:
# reverse() seems broken, using request.path but beware, it discards GET and POST info
return job_not_assigned(request, jobid, request.path)
def single_tracker(request, trackerid):
jt = get_api(request.user, request.jt)
try:
tracker = jt.get_tracker(trackerid)
except Exception, e:
raise PopupException(_('The tracker could not be contacted.'), detail=e)
def decorate(request, *args, **kwargs):
jobid = kwargs['job']
try:
job = get_api(request.user, request.jt).get_job(jobid=jobid)
except ApplicationNotRunning, e:
# reverse() seems broken, using request.path but beware, it discards GET and POST info
return job_not_assigned(request, jobid, request.path)
def single_tracker(request, trackerid):
jt = get_api(request.user, request.jt)
try:
tracker = jt.get_tracker(trackerid)
except Exception, e:
raise PopupException(_('The container disappears as soon as the job finishes.'), detail=e)
def kill_job(request, job):
if request.method != "POST":
raise Exception(_("kill_job may only be invoked with a POST (got a %(method)s).") % {'method': request.method})
if job.user != request.user.username and not request.user.is_superuser:
access_warn(request, _('Insufficient permission'))
raise MessageException(_("Permission denied. User %(username)s cannot delete user %(user)s's job.") % {'username': request.user.username, 'user': job.user})
job.kill()
cur_time = time.time()
api = get_api(request.user, request.jt)
while time.time() - cur_time < 15:
job = api.get_job(jobid=job.jobId)
if job.status not in ["RUNNING", "QUEUED"]:
if request.REQUEST.get("next"):
return HttpResponseRedirect(request.REQUEST.get("next"))
elif request.REQUEST.get("format") == "json":
return JsonResponse({'status': 0}, encoder=JSONEncoderForHTML)
else:
raise MessageException("Job Killed")
time.sleep(1)
raise Exception(_("Job did not appear as killed within 15 seconds."))
def jobs(request):
user = request.POST.get('user', request.user.username)
state = request.POST.get('state')
text = request.POST.get('text')
retired = request.POST.get('retired')
time_value = request.POST.get('time_value', 7)
time_unit = request.POST.get('time_unit', 'days')
if request.POST.get('format') == 'json':
try:
# Limit number of jobs to be 1000
jobs = get_api(request.user,
request.jt).get_jobs(user=request.user,
username=user,
state=state,
text=text,
retired=retired,
limit=1000,
time_value=int(time_value),
time_unit=time_unit)
except Exception as ex:
ex_message = str(ex)
if 'Connection refused' in ex_message or 'standby RM' in ex_message or 'Name or service not known' in ex_message:
response = {
'status':
1,
'message':
_('Resource Manager cannot be contacted or might be down.')
}
return JsonResponse(response)
elif 'Could not connect to' in ex_message:
raise PopupException(
_('Job Tracker cannot be contacted or might be down.'))
else:
raise PopupException(ex)
json_jobs = {
'jobs': [massage_job_for_json(job, request) for job in jobs],
}
return JsonResponse(json_jobs, encoder=JSONEncoderForHTML)
return render(
'jobs.mako', request, {
'request':
request,
'state_filter':
state,
'user_filter':
user,
'text_filter':
text,
'retired':
retired,
'filtered':
not (state == 'all' and user == '' and text == ''),
'is_yarn':
cluster.is_yarn(),
'hiveserver2_impersonation_enabled':
hiveserver2_impersonation_enabled()
})
def container(request, node_manager_http_address, containerid):
jt = get_api(request.user, request.jt)
try:
tracker = jt.get_tracker(node_manager_http_address, containerid)
except Exception, e:
# TODO: add a redirect of some kind
raise PopupException(_('The container disappears as soon as the job finishes.'), detail=e)
def test_yarn_configurations(user):
result = []
try:
from jobbrowser.api import get_api # Required for cluster HA testing
except Exception as e:
LOG.warn('Jobbrowser is disabled, skipping test_yarn_configurations')
return result
try:
get_api(user, None).get_jobs(user, username=user.username, state='all', text='')
except Exception as e:
msg = 'Failed to contact an active Resource Manager: %s' % e
LOG.exception(msg)
result.append(('Resource Manager', msg))
return result
def container(request, node_manager_http_address, containerid):
jt = get_api(request.user, request.jt)
try:
tracker = jt.get_tracker(node_manager_http_address, containerid)
except Exception, e:
# TODO: add a redirect of some kind
raise PopupException(_('The container disappears as soon as the job finishes.'), detail=e)
def single_tracker(request, trackerid):
jt = get_api(request.user, request.jt)
try:
tracker = jt.get_tracker(trackerid)
except Exception, e:
raise PopupException(
_('The container disappears as soon as the job finishes.'),
detail=e)
def get_job(request, job_id):
try:
job = get_api(request.user, request.jt).get_job(jobid=job_id)
except ApplicationNotRunning, e:
if e.job.get('state', '').lower() == 'accepted':
rm_api = resource_manager_api.get_resource_manager(request.user)
job = Application(e.job, rm_api)
else:
raise e # Job has not yet been accepted by RM
def get_job(request, job_id):
try:
job = get_api(request.user, request.jt).get_job(jobid=job_id)
except ApplicationNotRunning, e:
if e.job.get('state', '').lower() == 'accepted':
rm_api = resource_manager_api.get_resource_manager(request.user)
job = Application(e.job, rm_api)
else:
raise e # Job has not yet been accepted by RM
def job_not_assigned(request, jobid, path):
if request.GET.get('format') == 'json':
result = {'status': -1, 'message': ''}
try:
get_api(request.user, request.jt).get_job(jobid=jobid)
result['status'] = 0
except ApplicationNotRunning as e:
result['status'] = 1
except Exception as e:
result['message'] = _('Error polling job %s: %s') % (jobid, e)
return JsonResponse(result, encoder=JSONEncoderForHTML)
else:
return render('job_not_assigned.mako', request, {
'jobid': jobid,
'path': path
})
def decorate(request, *args, **kwargs):
jobid = kwargs['job']
try:
job = get_api(request.user, request.jt).get_job(jobid=jobid)
except Exception, e:
raise PopupException(
_('Could not find job %s. The job might not be running yet.') %
jobid,
detail=e)
def single_tracker(request, trackerid):
jt = get_api(request.user, request.jt)
try:
tracker = jt.get_tracker(trackerid)
except Exception as e:
raise PopupException(_('The tracker could not be contacted.'),
detail=e)
return render("tasktracker.mako", request, {'tracker': tracker})
def single_task_attempt_logs(request, job, taskid, attemptid):
jt = get_api(request.user, request.jt)
job_link = jt.get_job_link(job.jobId)
task = job_link.get_task(taskid)
try:
attempt = task.get_attempt(attemptid)
except (KeyError, RestException), e:
raise KeyError(_("Cannot find attempt '%(id)s' in task") % {'id': attemptid}, e)
def single_task_attempt(request, job, taskid, attemptid):
jt = get_api(request.user, request.jt)
job_link = jt.get_job_link(job.jobId)
task = job_link.get_task(taskid)
try:
attempt = task.get_attempt(attemptid)
except (KeyError, RestException), e:
raise PopupException(_("Cannot find attempt '%(id)s' in task") % {'id': attemptid}, e)
def single_task_attempt_logs(request, job, taskid, attemptid, offset=LOG_OFFSET_BYTES):
jt = get_api(request.user, request.jt)
job_link = jt.get_job_link(job.jobId)
task = job_link.get_task(taskid)
try:
attempt = task.get_attempt(attemptid)
except (KeyError, RestException), e:
raise KeyError(_("Cannot find attempt '%(id)s' in task") % {'id': attemptid}, e)
def decorate(request, *args, **kwargs):
jobid = kwargs['job']
try:
job = get_api(request.user, request.jt).get_job(jobid=jobid)
except ApplicationNotRunning, e:
if e.job.get('state', '').lower() == 'accepted' and 'kill' in request.path:
rm_api = resource_manager_api.get_resource_manager(request.user)
job = Application(e.job, rm_api)
else:
# reverse() seems broken, using request.path but beware, it discards GET and POST info
return job_not_assigned(request, jobid, request.path)
def decorate(request, *args, **kwargs):
jobid = kwargs['job']
try:
job = get_api(request.user, request.jt).get_job(jobid=jobid)
except ApplicationNotRunning, e:
if e.job.get('state', '').lower() == 'accepted' and 'kill' in request.path:
rm_api = resource_manager_api.get_resource_manager(request.user)
job = Application(e.job, rm_api)
else:
# reverse() seems broken, using request.path but beware, it discards GET and POST info
return job_not_assigned(request, jobid, request.path)
def kill_job_by_jt(request):
if request.method != "POST":
raise Exception("kill_job may only be invoked with a POST (got a %(method)s).") % dict(method=request.method)
job_id = request.POST['job_id']
# check job permission
try:
api = get_api(request.user, request.jt)
job = api.get_job(jobid=job_id)
except Exception, e:
return HttpResponse(json.dumps({"text": "Could not find job %s. The job might not be running yet." % job_id}))
def get_job(request, job_id):
try:
job = get_api(request.user, request.jt).get_job(jobid=job_id)
except ApplicationNotRunning, e:
if e.job.get('state',
'').lower() == 'accepted' and 'kill' in request.path:
rm_pool = resource_manager_api.get_resource_manager_pool()
rm_api = rm_pool.get(request.user.username)
job = Application(e.job, rm_api)
rm_pool.put(rm_api)
else:
# reverse() seems broken, using request.path but beware, it discards GET and POST info
return job_not_assigned(request, job_id, request.path)
def jobs(request):
user = request.GET.get('user', request.user.username)
state = request.GET.get('state')
text = request.GET.get('text')
retired = request.GET.get('retired')
jobs = get_api(request.user, request.jt).get_jobs(user=request.user, username=user, state=state, text=text, retired=retired)
return render('jobs.mako', request, {
'jobs': jobs,
'request': request,
'state_filter': state,
'user_filter': user,
'text_filter': text,
'retired': retired,
'filtered': not (state == 'all' and user == '' and text == '')
})
def single_task_attempt(request, job, taskid, attemptid):
jt = get_api(request.user, request.jt)
job_link = jt.get_job_link(job.jobId)
task = job_link.get_task(taskid)
try:
attempt = task.get_attempt(attemptid)
except (KeyError, RestException) as e:
raise PopupException(
_("Cannot find attempt '%(id)s' in task") % {'id': attemptid}, e)
return render("attempt.mako", request, {
"attempt": attempt,
"taskid": taskid,
"joblnk": job_link,
"task": task
})
def get_job(request, job_id):
try:
job = get_api(request.user, request.jt).get_job(jobid=job_id)
except ApplicationNotRunning as e:
if e.job.get('state', '').lower() == 'accepted':
rm_api = resource_manager_api.get_resource_manager(request.user)
job = Application(e.job, rm_api)
else:
raise e # Job has not yet been accepted by RM
except JobExpired as e:
raise PopupException(
_('Job %s has expired.') % job_id,
detail=_('Cannot be found on the History Server.'))
except Exception as e:
msg = 'Could not find job %s.'
LOG.exception(msg % job_id)
raise PopupException(_(msg) % job_id, detail=e)
return job
def kill_job_by_jt(request):
if request.method != "POST":
raise Exception(
"kill_job may only be invoked with a POST (got a %(method)s)."
) % dict(method=request.method)
job_id = request.POST['job_id']
# check job permission
try:
api = get_api(request.user, request.jt)
job = api.get_job(jobid=job_id)
except Exception, e:
return HttpResponse(
json.dumps({
"text":
"Could not find job %s. The job might not be running yet." %
job_id
}))
def jobs(request):
user = request.GET.get('user', request.user.username)
state = request.GET.get('state')
text = request.GET.get('text')
retired = request.GET.get('retired')
if request.GET.get('format') == 'json':
try:
jobs = get_api(request.user, request.jt).get_jobs(user=request.user, username=user, state=state, text=text, retired=retired)
except Exception, ex:
ex_message = str(ex)
if 'Connection refused' in ex_message or 'standby RM' in ex_message:
raise PopupException(_('Resource Manager cannot be contacted or might be down.'))
elif 'Could not connect to' in ex_message:
raise PopupException(_('Job Tracker cannot be contacted or might be down.'))
else:
raise ex
json_jobs = [massage_job_for_json(job, request) for job in jobs]
return HttpResponse(encode_json_for_js(json_jobs), mimetype="application/json")
def jobs(request):
user = request.GET.get('user', request.user.username)
state = request.GET.get('state')
text = request.GET.get('text')
retired = request.GET.get('retired')
if request.GET.get('format') == 'json':
jobs = get_api(request.user, request.jt).get_jobs(user=request.user, username=user, state=state, text=text, retired=retired)
json_jobs = [massage_job_for_json(job, request) for job in jobs]
return HttpResponse(encode_json_for_js(json_jobs), mimetype="application/json")
return render('jobs.mako', request, {
'request': request,
'state_filter': state,
'user_filter': user,
'text_filter': text,
'retired': retired,
'filtered': not (state == 'all' and user == '' and text == '')
})
def jobs(request):
user = request.GET.get('user', request.user.username)
state = request.GET.get('state')
text = request.GET.get('text')
retired = request.GET.get('retired')
if request.GET.get('format') == 'json':
jobs = get_api(request.user, request.jt).get_jobs(user=request.user, username=user, state=state, text=text, retired=retired)
json_jobs = [massage_job_for_json(job, request) for job in jobs]
return HttpResponse(encode_json_for_js(json_jobs), mimetype="application/json")
return render('jobs.mako', request, {
'request': request,
'state_filter': state,
'user_filter': user,
'text_filter': text,
'retired': retired,
'filtered': not (state == 'all' and user == '' and text == '')
})
def tasks(request, job):
"""
We get here from /jobs/job/tasks?filterargs, with the options being:
page=<n> - Controls pagination. Defaults to 1.
tasktype=<type> - Type can be one of hadoop.job_tracker.VALID_TASK_TYPES
("map", "reduce", "job_cleanup", "job_setup")
taskstate=<state> - State can be one of hadoop.job_tracker.VALID_TASK_STATES
("succeeded", "failed", "running", "pending", "killed")
tasktext=<text> - Where <text> is a string matching info on the task
"""
ttypes = request.GET.get("tasktype")
tstates = request.GET.get("taskstate")
ttext = request.GET.get("tasktext")
pagenum = int(request.GET.get("page", 1))
pagenum = pagenum > 0 and pagenum or 1
filters = {
"task_types": ttypes and set(ttypes.split(",")) or None,
"task_states": tstates and set(tstates.split(",")) or None,
"task_text": ttext,
"pagenum": pagenum,
}
jt = get_api(request.user, request.jt)
task_list = jt.get_tasks(job.jobId, **filters)
filter_params = copy_query_dict(request.GET, ("tasktype", "taskstate", "tasktext")).urlencode()
return render(
"tasks.mako",
request,
{
"request": request,
"filter_params": filter_params,
"job": job,
"task_list": task_list,
"tasktype": ttypes,
"taskstate": tstates,
"tasktext": ttext,
},
)
def tasks(request, job):
"""
We get here from /jobs/job/tasks?filterargs, with the options being:
page=<n> - Controls pagination. Defaults to 1.
tasktype=<type> - Type can be one of hadoop.job_tracker.VALID_TASK_TYPES
("map", "reduce", "job_cleanup", "job_setup")
taskstate=<state> - State can be one of hadoop.job_tracker.VALID_TASK_STATES
("succeeded", "failed", "running", "pending", "killed")
tasktext=<text> - Where <text> is a string matching info on the task
"""
ttypes = request.GET.get('tasktype')
tstates = request.GET.get('taskstate')
ttext = request.GET.get('tasktext')
pagenum = int(request.GET.get('page', 1))
pagenum = pagenum > 0 and pagenum or 1
filters = {
'task_types': ttypes and set(ttypes.split(',')) or None,
'task_states': tstates and set(tstates.split(',')) or None,
'task_text': ttext,
'pagenum': pagenum,
}
jt = get_api(request.user, request.jt)
task_list = jt.get_tasks(job.jobId, **filters)
page = jt.paginate_task(task_list, pagenum)
filter_params = copy_query_dict(
request.GET, ('tasktype', 'taskstate', 'tasktext')).urlencode()
return render(
"tasks.mako", request, {
'request': request,
'filter_params': filter_params,
'job': job,
'page': page,
'tasktype': ttypes,
'taskstate': tstates,
'tasktext': ttext
})
def job_attempt_logs_json(request, job, attempt_index=0, name="syslog", offset=LOG_OFFSET_BYTES):
"""For async log retrieval as Yarn servers are very slow"""
log_link = None
response = {"status": -1}
try:
jt = get_api(request.user, request.jt)
app = jt.get_application(job.jobId)
if app["applicationType"] == "MAPREDUCE":
if app["finalStatus"] in ("SUCCEEDED", "FAILED", "KILLED"):
attempt_index = int(attempt_index)
attempt = job.job_attempts["jobAttempt"][attempt_index]
log_link = attempt["logsLink"]
# Reformat log link to use YARN RM, replace node addr with node ID addr
log_link = log_link.replace(attempt["nodeHttpAddress"], attempt["nodeId"])
elif app["state"] == "RUNNING":
log_link = app["amContainerLogs"]
except (KeyError, RestException), e:
raise KeyError(_("Cannot find job attempt '%(id)s'.") % {"id": job.jobId}, e)
def job_attempt_logs_json(request, job, attempt_index=0, name='syslog', offset=LOG_OFFSET_BYTES):
"""For async log retrieval as Yarn servers are very slow"""
log_link = None
response = {'status': -1}
try:
jt = get_api(request.user, request.jt)
app = jt.get_application(job.jobId)
if app['applicationType'] == 'MAPREDUCE':
if app['finalStatus'] in ('SUCCEEDED', 'FAILED', 'KILLED'):
attempt_index = int(attempt_index)
attempt = job.job_attempts['jobAttempt'][attempt_index]
log_link = attempt['logsLink']
# Reformat log link to use YARN RM, replace node addr with node ID addr
log_link = log_link.replace(attempt['nodeHttpAddress'], attempt['nodeId'])
elif app['state'] == 'RUNNING':
log_link = app['amContainerLogs']
except (KeyError, RestException), e:
raise KeyError(_("Cannot find job attempt '%(id)s'.") % {'id': job.jobId}, e)
def jobs(request):
user = request.GET.get("user", request.user.username)
state = request.GET.get("state")
text = request.GET.get("text")
retired = request.GET.get("retired")
if request.GET.get("format") == "json":
try:
# Limit number of jobs to be 10,000
jobs = get_api(request.user, request.jt).get_jobs(
user=request.user, username=user, state=state, text=text, retired=retired, limit=10000
)
except Exception, ex:
ex_message = str(ex)
if "Connection refused" in ex_message or "standby RM" in ex_message:
raise PopupException(_("Resource Manager cannot be contacted or might be down."))
elif "Could not connect to" in ex_message:
raise PopupException(_("Job Tracker cannot be contacted or might be down."))
else:
raise ex
json_jobs = {"jobs": [massage_job_for_json(job, request) for job in jobs]}
return JsonResponse(json_jobs, encoder=JSONEncoderForHTML)
def jobs(request):
user = request.GET.get('user', request.user.username)
state = request.GET.get('state')
text = request.GET.get('text')
retired = request.GET.get('retired')
jobs = get_api(request.user, request.jt).get_jobs(user=request.user,
username=user,
state=state,
text=text,
retired=retired)
return render(
'jobs.mako', request, {
'jobs': jobs,
'request': request,
'state_filter': state,
'user_filter': user,
'text_filter': text,
'retired': retired,
'filtered': not (state == 'all' and user == '' and text == '')
})
@access_log_level(logging.WARN)
@check_job_permission
def kill_job(request, job):
if request.method != "POST":
raise Exception(
_("kill_job may only be invoked with a POST (got a %(method)s).") %
{'method': request.method})
try:
job.kill()
except Exception, e:
LOG.exception('Killing job')
raise PopupException(e)
cur_time = time.time()
api = get_api(request.user, request.jt)
while time.time() - cur_time < 15:
try:
job = api.get_job(jobid=job.jobId)
except Exception, e:
LOG.warn('Failed to get job with ID %s: %s' % (job.jobId, e))
else:
if job.status not in ["RUNNING", "QUEUED"]:
if request.REQUEST.get("next"):
return HttpResponseRedirect(request.REQUEST.get("next"))
elif request.REQUEST.get("format") == "json":
return JsonResponse({'status': 0},
encoder=JSONEncoderForHTML)
else:
raise MessageException("Job Killed")
res.extend(test_yarn_configurations(user))
for name in YARN_CLUSTERS.keys():
cluster = YARN_CLUSTERS[name]
if cluster.SUBMIT_TO.get():
submit_to.append('yarn_clusters.' + name)
if not submit_to:
res.append(("hadoop", "Please designate one of the MapReduce or "
"Yarn clusters with `submit_to=true' in order to run jobs."))
return res
def test_yarn_configurations(user):
result = []
try:
from jobbrowser.api import get_api # Required for cluster HA testing
except Exception, e:
LOG.warn('Jobbrowser is disabled, skipping test_yarn_configurations')
return result
try:
get_api(user, None).get_jobs(user, username=user.username, state='all', text='')
except Exception, e:
msg = 'Failed to contact an active Resource Manager: %s' % e
LOG.exception(msg)
result.append(('Resource Manager', msg))
return result
def decorate(request, *args, **kwargs):
jobid = kwargs['job']
try:
job = get_api(request.user, request.jt).get_job(jobid=jobid)
except Exception, e:
raise PopupException(_('Could not find job %s. The job might not be running yet.') % jobid, detail=e)
def single_task_attempt_logs(request,
job,
taskid,
attemptid,
offset=LOG_OFFSET_BYTES):
jt = get_api(request.user, request.jt)
job_link = jt.get_job_link(job.jobId)
task = job_link.get_task(taskid)
try:
attempt = task.get_attempt(attemptid)
except (KeyError, RestException) as e:
raise KeyError(
_("Cannot find attempt '%(id)s' in task") % {'id': attemptid}, e)
first_log_tab = 0
try:
# Add a diagnostic log
if hasattr(task, 'job') and hasattr(task.job, 'diagnostics'):
diagnostic_log = task.job.diagnostics
elif job_link.is_mr2:
diagnostic_log = attempt.diagnostics
else:
diagnostic_log = ", ".join(task.diagnosticMap[attempt.attemptId])
logs = [diagnostic_log]
# Add remaining logs
logs += [
section.strip() for section in attempt.get_task_log(offset=offset)
]
log_tab = [i for i, log in enumerate(logs) if log]
if log_tab:
first_log_tab = log_tab[0]
except urllib.error.URLError:
logs = [_("Failed to retrieve log. TaskTracker not ready.")] * 4
context = {
"attempt": attempt,
"taskid": taskid,
"joblnk": job_link,
"task": task,
"logs": logs,
"logs_list": attempt.get_log_list(),
"first_log_tab": first_log_tab,
}
if request.GET.get('format') == 'python':
return context
else:
context['logs'] = [
LinkJobLogs._make_links(log) for i, log in enumerate(logs)
]
if request.GET.get('format') == 'json':
response = {
"logs": context['logs'],
"logsList": context['logs_list'],
"isRunning": job.status.lower() in ('running', 'pending', 'prep')
}
return JsonResponse(response)
else:
return render("attempt_logs.mako", request, context)
def single_tracker(request, trackerid):
jt = get_api(request.user, request.jt)
tracker = jt.get_tracker(trackerid)
return render("tasktracker.mako", request, {'tracker': tracker})
def single_tracker(request, trackerid):
jt = get_api(request.user, request.jt)
tracker = jt.get_tracker(trackerid)
return render("tasktracker.mako", request, {'tracker':tracker})
def job_attempt_logs_json(request,
job,
attempt_index=0,
name='syslog',
offset=LOG_OFFSET_BYTES,
is_embeddable=False):
"""For async log retrieval as Yarn servers are very slow"""
log_link = None
response = {'status': -1}
try:
jt = get_api(request.user, request.jt)
app = jt.get_application(job.jobId)
if app['applicationType'] == 'MAPREDUCE':
if app['finalStatus'] in ('SUCCEEDED', 'FAILED', 'KILLED'):
attempt_index = int(attempt_index)
if not job.job_attempts['jobAttempt']:
response = {'status': 0, 'log': _('Job has no tasks')}
else:
attempt = job.job_attempts['jobAttempt'][attempt_index]
log_link = attempt['logsLink']
# Reformat log link to use YARN RM, replace node addr with node ID addr
log_link = log_link.replace(attempt['nodeHttpAddress'],
attempt['nodeId'])
elif app['state'] == 'RUNNING':
log_link = app['amContainerLogs']
elif app.get('amContainerLogs'):
log_link = app.get('amContainerLogs')
except (KeyError, RestException) as e:
raise KeyError(
_("Cannot find job attempt '%(id)s'.") % {'id': job.jobId}, e)
except Exception as e:
raise Exception(
_("Failed to get application for job %s: %s") % (job.jobId, e))
if log_link:
link = '/%s/' % name
params = {'doAs': request.user.username}
if offset != 0:
params['start'] = offset
root = Resource(get_log_client(log_link),
urllib.parse.urlsplit(log_link)[2],
urlencode=False)
api_resp = None
try:
api_resp = root.get(link, params=params)
log = html.fromstring(api_resp, parser=html.HTMLParser()).xpath(
'/html/body/table/tbody/tr/td[2]')[0].text_content()
response['status'] = 0
response['log'] = LinkJobLogs._make_hdfs_links(log, is_embeddable)
except Exception as e:
response['log'] = _('Failed to retrieve log: %s' % e)
try:
debug_info = '\nLog Link: %s' % log_link
if api_resp:
debug_info += '\nHTML Response: %s' % response
response['debug'] = debug_info
LOG.error(debug_info)
except:
LOG.exception('failed to create debug info')
return JsonResponse(response)
submit_to.append('yarn_clusters.' + name)
if not submit_to:
res.append(
("hadoop", "Please designate one of the MapReduce or "
"Yarn clusters with `submit_to=true' in order to run jobs."))
return res
def test_yarn_configurations(user):
result = []
try:
from jobbrowser.api import get_api # Required for cluster HA testing
except Exception, e:
LOG.warn('Jobbrowser is disabled, skipping test_yarn_configurations')
return result
try:
get_api(user, None).get_jobs(user,
username=user.username,
state='all',
text='')
except Exception, e:
msg = 'Failed to contact an active Resource Manager: %s' % e
LOG.exception(msg)
result.append(('Resource Manager', msg))
return result
def kill_job(request, job):
if request.method != "POST":
raise Exception(_("kill_job may only be invoked with a POST (got a %(method)s).") % {'method': request.method})
if job.user != request.user.username and not request.user.is_superuser:
access_warn(request, _('Insufficient permission'))
raise MessageException(_("Permission denied. User %(username)s cannot delete user %(user)s's job.") % {'username': request.user.username, 'user': job.user})
try:
job.kill()
except Exception, e:
LOG.exception('Killing job')
raise PopupException(e)
cur_time = time.time()
api = get_api(request.user, request.jt)
while time.time() - cur_time < 15:
try:
job = api.get_job(jobid=job.jobId)
except Exception, e:
LOG.warn('Failed to get job with ID %s: %s' % (job.jobId, e))
else:
if job.status not in ["RUNNING", "QUEUED"]:
if request.REQUEST.get("next"):
return HttpResponseRedirect(request.REQUEST.get("next"))
elif request.REQUEST.get("format") == "json":
return JsonResponse({'status': 0}, encoder=JSONEncoderForHTML)
else:
raise MessageException("Job Killed")
time.sleep(1)
