def monitor():
try:
window_size = arg_cardinal('window')
except ValueError as ex:
return bad_request(str(ex))
try:
cursor = dbcursor_query(
"""SELECT ppf, lower(times), upper(times), task, run,
state_enum, state_display, task_json,
task_cli FROM schedule_monitor(%s)""",
[window_size])
except Exception as ex:
log.exception()
return error(str(ex))
result = []
base_url = pscheduler.api_url_hostport(server_netloc())
for row in cursor:
task_href = "%s/tasks/%s" % (base_url, row[2])
run_href = "%s/runs/%s" % (task_href, row[3])
run = {
"ppf": row[0],
"start-time": pscheduler.datetime_as_iso8601(row[1]),
"end-time": pscheduler.datetime_as_iso8601(row[2]),
"href": run_href,
"result-href": "%s/result" % run_href,
"state": row[5],
"state-display": row[6],
"task": row[7],
"cli": row[8]
}
run["task"]["href"] = task_href
result.append(run)
return ok_json(result)
def tasks():
if request.method == 'GET':
where_clause = "TRUE"
args = []
try:
json_query = arg_json("json")
except ValueError as ex:
return bad_request(str(ex))
if json_query is not None:
where_clause += " AND task.json @> %s"
args.append(request.args.get("json"))
where_clause += " ORDER BY added"
try:
tasks = __tasks_get_filtered(request.base_url,
where_clause=where_clause,
args=args,
expanded=is_expanded(),
detail=arg_boolean("detail"),
single=False)
except Exception as ex:
return error(str(ex))
return ok_json(tasks)
elif request.method == 'POST':
try:
task = pscheduler.json_load(request.data, max_schema=1)
except ValueError as ex:
return bad_request("Invalid task specification: %s" % (str(ex)))
# Validate the JSON against a TaskSpecification
# TODO: Figure out how to do this without the intermediate object
valid, message = pscheduler.json_validate({"": task}, {
"type": "object",
"properties": {
"": {
"$ref": "#/pScheduler/TaskSpecification"
}
},
"required": [""]
})
if not valid:
return bad_request("Invalid task specification: %s" % (message))
# See if the test spec is valid
try:
returncode, stdout, stderr = pscheduler.run_program(
[
"pscheduler", "internal", "invoke", "test",
task['test']['type'], "spec-is-valid"
],
stdin=pscheduler.json_dump(task['test']['spec']))
if returncode != 0:
return error("Unable to validate test spec: %s" % (stderr))
validate_json = pscheduler.json_load(stdout, max_schema=1)
if not validate_json["valid"]:
return bad_request(
"Invalid test specification: %s" %
(validate_json.get("error", "Unspecified error")))
except Exception as ex:
return error("Unable to validate test spec: " + str(ex))
log.debug("Validated test: %s", pscheduler.json_dump(task['test']))
# Reject tasks that have archive specs that use transforms.
# See ticket #330.
try:
for archive in task['archives']:
if "transform" in archive:
return bad_request(
"Use of transforms in archives is not yet supported.")
except KeyError:
pass # Not there
# Find the participants
try:
# HACK: BWCTLBC
if "lead-bind" in task:
lead_bind_env = {
"PSCHEDULER_LEAD_BIND_HACK": task["lead-bind"]
}
else:
lead_bind_env = None
returncode, stdout, stderr = pscheduler.run_program(
[
"pscheduler", "internal", "invoke", "test",
task['test']['type'], "participants"
],
stdin=pscheduler.json_dump(task['test']['spec']),
timeout=5,
env_add=lead_bind_env)
if returncode != 0:
return error("Unable to determine participants: " + stderr)
participants = [
host if host is not None else server_netloc() for host in
pscheduler.json_load(stdout, max_schema=1)["participants"]
]
except Exception as ex:
return error("Exception while determining participants: " +
str(ex))
nparticipants = len(participants)
# TODO: The participants must be unique. This should be
# verified by fetching the host name from each one.
#
# TOOL SELECTION
#
lead_bind = task.get("lead-bind", None)
# TODO: Need to provide for tool being specified by the task
# package.
tools = []
tool_params = {"test": pscheduler.json_dump(task["test"])}
# HACK: BWCTLBC
if lead_bind is not None:
log.debug("Using lead bind of %s" % str(lead_bind))
tool_params["lead-bind"] = lead_bind
for participant_no in range(0, len(participants)):
participant = participants[participant_no]
try:
# Make sure the other participants are running pScheduler
participant_api = pscheduler.api_url_hostport(participant)
log.debug("Pinging %s" % (participant))
status, result = pscheduler.url_get(participant_api,
throw=False,
timeout=10,
bind=lead_bind)
if status == 400:
raise TaskPostingException(result)
elif status in [ 202, 204, 205, 206, 207, 208, 226,
300, 301, 302, 303, 304, 205, 306, 307, 308 ] \
or ( (status >= 400) and (status <=499) ):
raise TaskPostingException(
"Host is not running pScheduler")
elif status != 200:
raise TaskPostingException("returned status %d: %s" %
(status, result))
# TODO: This will fail with a very large test spec.
status, result = pscheduler.url_get("%s/tools" %
(participant_api),
params=tool_params,
throw=False,
bind=lead_bind)
if status != 200:
raise TaskPostingException("%d: %s" % (status, result))
tools.append(result)
except TaskPostingException as ex:
return error("Error getting tools from %s: %s" \
% (participant, str(ex)))
log.debug("Participant %s offers tools %s", participant, result)
if len(tools) != nparticipants:
return error("Didn't get a full set of tool responses")
if "tools" in task:
tool = pick_tool(tools, pick_from=task['tools'])
else:
tool = pick_tool(tools)
if tool is None:
# TODO: This could stand some additional diagnostics.
return no_can_do(
"Couldn't find a tool in common among the participants.")
task['tool'] = tool
#
# TASK CREATION
#
tasks_posted = []
# Evaluate the task against the limits and reject the request
# if it doesn't pass.
log.debug("Checking limits on %s", task["test"])
(processor, whynot) = limitprocessor()
if processor is None:
log.debug("Limit processor is not initialized. %s", whynot)
return no_can_do("Limit processor is not initialized: %s" % whynot)
hints = request_hints()
hints_data = pscheduler.json_dump(hints)
log.debug("Processor = %s" % processor)
passed, limits_passed, diags = processor.process(task["test"], hints)
if not passed:
return forbidden("Task forbidden by limits:\n" + diags)
# Post the lead with the local database, which also assigns
# its UUID. Make it disabled so the scheduler doesn't try to
# do anything with it until the task has been submitted to all
# of the other participants.
try:
cursor = dbcursor_query(
"SELECT * FROM api_task_post(%s, %s, %s, %s, 0, NULL, FALSE)",
[
pscheduler.json_dump(task), participants, hints_data,
pscheduler.json_dump(limits_passed)
],
onerow=True)
except Exception as ex:
return error(str(ex.diag.message_primary))
if cursor.rowcount == 0:
return error("Task post failed; poster returned nothing.")
task_uuid = cursor.fetchone()[0]
log.debug("Tasked lead, UUID %s", task_uuid)
# Other participants get the UUID and participant list forced upon them.
task["participants"] = participants
task_data = pscheduler.json_dump(task)
for participant in range(1, nparticipants):
part_name = participants[participant]
log.debug("Tasking participant %s", part_name)
try:
# Post the task
log.debug("Tasking %d@%s: %s", participant, part_name,
task_data)
post_url = pscheduler.api_url_hostport(part_name,
'tasks/' + task_uuid)
log.debug("Posting task to %s", post_url)
status, result = pscheduler.url_post(
post_url,
params={'participant': participant},
data=task_data,
bind=lead_bind,
json=False,
throw=False)
log.debug("Remote returned %d: %s", status, result)
if status != 200:
raise TaskPostingException(
"Unable to post task to %s: %s" % (part_name, result))
tasks_posted.append(result)
# Fetch the task's details and add the list of limits
# passed to our own.
status, result = pscheduler.url_get(post_url,
params={"detail": True},
bind=lead_bind,
throw=False)
if status != 200:
raise TaskPostingException(
"Unable to fetch posted task from %s: %s" %
(part_name, result))
log.debug("Fetched %s", result)
try:
details = result["detail"]["spec-limits-passed"]
log.debug("Details from %s: %s", post_url, details)
limits_passed.extend(details)
except KeyError:
pass
except TaskPostingException as ex:
# Disable the task locally and let it get rid of the
# other participants.
posted_to = "%s/%s" % (request.url, task_uuid)
parsed = list(urlparse.urlsplit(posted_to))
parsed[1] = "%s"
template = urlparse.urlunsplit(parsed)
try:
dbcursor_query("SELECT api_task_disable(%s, %s)",
[task_uuid, template])
except Exception:
log.exception()
return error("Error while tasking %s: %s" % (part_name, ex))
# Update the list of limits passed in the local database
# TODO: How do the other participants know about this?
log.debug("Limits passed: %s", limits_passed)
try:
cursor = dbcursor_query(
"UPDATE task SET limits_passed = %s::JSON WHERE uuid = %s",
[pscheduler.json_dump(limits_passed), task_uuid])
except Exception as ex:
return error(str(ex.diag.message_primary))
# Enable the task so the scheduler will schedule it.
try:
dbcursor_query("SELECT api_task_enable(%s)", [task_uuid])
except Exception as ex:
log.exception()
return error("Failed to enable task %s. See system logs." %
task_uuid)
log.debug("Task enabled for scheduling.")
return ok_json("%s/%s" % (request.base_url, task_uuid))
else:
return not_allowed()
def tasks():
if request.method == 'GET':
where_clause = "TRUE"
args = []
try:
json_query = arg_json("json")
except ValueError as ex:
return bad_request(str(ex))
if json_query is not None:
where_clause += " AND task.json_detail @> %s"
args.append(request.args.get("json"))
where_clause += " ORDER BY added"
tasks = __tasks_get_filtered(request.base_url,
where_clause=where_clause,
args=args,
expanded=is_expanded(),
detail=arg_boolean("detail"),
single=False)
return ok_json(tasks)
elif request.method == 'POST':
data = request.data.decode("ascii")
try:
task = pscheduler.json_load(data, max_schema=3)
except ValueError as ex:
return bad_request("Invalid task specification: %s" % (str(ex)))
# Validate the JSON against a TaskSpecification
# TODO: Figure out how to do this without the intermediate object
valid, message = pscheduler.json_validate({"": task}, {
"type": "object",
"properties": {
"": {
"$ref": "#/pScheduler/TaskSpecification"
}
},
"required": [""]
})
if not valid:
return bad_request("Invalid task specification: %s" % (message))
# See if the test spec is valid
try:
returncode, stdout, stderr = pscheduler.plugin_invoke(
"test",
task['test']['type'],
"spec-is-valid",
stdin=pscheduler.json_dump(task['test']['spec']))
if returncode != 0:
return error("Unable to validate test spec: %s" % (stderr))
validate_json = pscheduler.json_load(stdout, max_schema=1)
if not validate_json["valid"]:
return bad_request(
"Invalid test specification: %s" %
(validate_json.get("error", "Unspecified error")))
except Exception as ex:
return error("Unable to validate test spec: " + str(ex))
log.debug("Validated test: %s", pscheduler.json_dump(task['test']))
# Validate the schedule
try:
cron = crontab.CronTab(task["schedule"]["repeat-cron"])
except (AttributeError, ValueError):
return error("Cron repeat specification is invalid.")
except KeyError:
pass
# Validate the archives
for archive in task.get("archives", []):
# Data
try:
returncode, stdout, stderr = pscheduler.plugin_invoke(
"archiver",
archive["archiver"],
"data-is-valid",
stdin=pscheduler.json_dump(archive["data"]),
)
if returncode != 0:
return error("Unable to validate archive spec: %s" %
(stderr))
except Exception as ex:
return error("Unable to validate test spec: " + str(ex))
try:
returned_json = pscheduler.json_load(stdout)
if not returned_json["valid"]:
return bad_request("Invalid archiver data: %s" %
(returned_json["error"]))
except Exception as ex:
return error("Internal probelm validating archiver data: %s" %
(str(ex)))
# Transform, if there was one.
if "transform" in archive:
transform = archive["transform"]
try:
_ = pscheduler.JQFilter(filter_spec=transform["script"],
args=transform.get("args", {}))
except ValueError as ex:
return error("Invalid transform: %s" % (str(ex)))
# Validate the lead binding if there was one.
lead_bind = task.get("lead-bind", None)
if lead_bind is not None \
and (pscheduler.address_interface(lead_bind) is None):
return bad_request("Lead bind '%s' is not on this host" %
(lead_bind))
# Evaluate the task against the limits and reject the request
# if it doesn't pass. We do this early so anything else in
# the process gets any rewrites.
log.debug("Checking limits on %s", task)
(processor, whynot) = limitprocessor()
if processor is None:
log.debug("Limit processor is not initialized. %s", whynot)
return no_can_do("Limit processor is not initialized: %s" % whynot)
hints, error_response = request_hints()
if hints is None:
log.debug("Can't come up with valid hints for lead task limits.")
return error_response
hints_data = pscheduler.json_dump(hints)
log.debug("Processor = %s" % processor)
passed, limits_passed, diags, new_task, _priority \
= processor.process(task, hints)
if not passed:
return forbidden("Task forbidden by limits:\n" + diags)
if new_task is not None:
try:
task = new_task
returncode, stdout, stderr = pscheduler.plugin_invoke(
"test",
task['test']['type'],
"spec-is-valid",
stdin=pscheduler.json_dump(task["test"]["spec"]))
if returncode != 0:
return error(
"Failed to validate rewritten test specification: %s" %
(stderr))
validate_json = pscheduler.json_load(stdout, max_schema=1)
if not validate_json["valid"]:
return bad_request(
"Rewritten test specification is invalid: %s" %
(validate_json.get("error", "Unspecified error")))
except Exception as ex:
return error(
"Unable to validate rewritten test specification: " +
str(ex))
# Find the participants
try:
returncode, stdout, stderr = pscheduler.plugin_invoke(
"test",
task['test']['type'],
"participants",
stdin=pscheduler.json_dump(task['test']['spec']),
timeout=5)
if returncode != 0:
return error("Unable to determine participants: " + stderr)
participants = [
host if host is not None else server_netloc() for host in
pscheduler.json_load(stdout, max_schema=1)["participants"]
]
except Exception as ex:
return error("Exception while determining participants: " +
str(ex))
nparticipants = len(participants)
# TODO: The participants must be unique. This should be
# verified by fetching the host name from each one.
#
# TOOL SELECTION
#
# TODO: Need to provide for tool being specified by the task
# package.
tools = []
tool_params = {"test": pscheduler.json_dump(task["test"])}
tool_offers = {}
for participant_no in range(0, len(participants)):
participant = participants[participant_no]
try:
# Make sure the other participants are running pScheduler
participant_api = pscheduler.api_url_hostport(participant)
log.debug("Pinging %s" % (participant))
status, result = pscheduler.url_get(participant_api,
throw=False,
timeout=10,
bind=lead_bind)
if status == 400:
raise TaskPostingException(result)
elif status in [ 202, 204, 205, 206, 207, 208, 226,
300, 301, 302, 303, 304, 205, 306, 307, 308 ] \
or ( (status >= 400) and (status <=499) ):
raise TaskPostingException(
"Host is not running pScheduler")
elif status != 200:
raise TaskPostingException("returned status %d: %s" %
(status, result))
# TODO: This will fail with a very large test spec.
status, result = pscheduler.url_get("%s/tools" %
(participant_api),
params=tool_params,
throw=False,
bind=lead_bind)
if status != 200:
raise TaskPostingException("%d: %s" % (status, result))
tools.append(result)
except TaskPostingException as ex:
return error("Error getting tools from %s: %s" \
% (participant, str(ex)))
log.debug("Participant %s offers tools %s", participant, result)
tool_offers[participant] = result
if len(tools) != nparticipants:
return error("Didn't get a full set of tool responses")
if "tools" in task:
tool = pick_tool(tools, pick_from=task['tools'])
else:
tool = pick_tool(tools)
# Complain if no usable tool was found
if tool is None:
offers = []
for participant in participants:
participant_offers = tool_offers.get(participant,
[{
"name": "nothing"
}])
if participant_offers is not None:
offer_set = [offer["name"] for offer in participant_offers]
else:
offer_set = ["nothing"]
offers.append("%s offered %s" %
(participant, ", ".join(offer_set)))
return no_can_do("No tool in common among the participants: %s." %
("; ".join(offers)))
task['tool'] = tool
#
# TASK CREATION
#
tasks_posted = []
# Post the lead with the local database, which also assigns
# its UUID. Make it disabled so the scheduler doesn't try to
# do anything with it until the task has been submitted to all
# of the other participants.
cursor = dbcursor_query(
"SELECT * FROM api_task_post(%s, %s, %s, %s, 0, %s, NULL, FALSE, %s)",
[
pscheduler.json_dump(task), participants, hints_data,
pscheduler.json_dump(limits_passed),
task.get("priority", None), diags
],
onerow=True)
if cursor.rowcount == 0:
return error("Task post failed; poster returned nothing.")
task_uuid = cursor.fetchone()[0]
log.debug("Tasked lead, UUID %s", task_uuid)
# Other participants get the UUID and participant list forced upon them.
task["participants"] = participants
task_params = {"key": task["_key"]} if "_key" in task else {}
for participant in range(1, nparticipants):
part_name = participants[participant]
log.debug("Tasking participant %s", part_name)
try:
# Post the task
log.debug("Tasking %d@%s: %s", participant, part_name, task)
post_url = pscheduler.api_url_hostport(part_name,
'tasks/' + task_uuid)
task_params["participant"] = participant
log.debug("Posting task to %s", post_url)
status, result = pscheduler.url_post(post_url,
params=task_params,
data=task,
bind=lead_bind,
json=False,
throw=False)
log.debug("Remote returned %d: %s", status, result)
if status != 200:
raise TaskPostingException(
"Unable to post task to %s: %s" % (part_name, result))
tasks_posted.append(result)
# Fetch the task's details and add the list of limits
# passed to our own.
status, result = pscheduler.url_get(post_url,
params={"detail": True},
bind=lead_bind,
throw=False)
if status != 200:
raise TaskPostingException(
"Unable to fetch posted task from %s: %s" %
(part_name, result))
log.debug("Fetched %s", result)
try:
details = result["detail"]["spec-limits-passed"]
log.debug("Details from %s: %s", post_url, details)
limits_passed.extend(details)
except KeyError:
pass
except TaskPostingException as ex:
# Disable the task locally and let it get rid of the
# other participants.
posted_to = "%s/%s" % (request.url, task_uuid)
parsed = list(urllib.parse.urlsplit(posted_to))
parsed[1] = "%s"
template = urllib.parse.urlunsplit(parsed)
try:
dbcursor_query("SELECT api_task_disable(%s, %s)",
[task_uuid, template])
except Exception:
log.exception()
return error("Error while tasking %s: %s" % (part_name, ex))
# Update the list of limits passed in the local database
# TODO: How do the other participants know about this?
log.debug("Limits passed: %s", limits_passed)
cursor = dbcursor_query(
"UPDATE task SET limits_passed = %s::JSON WHERE uuid = %s",
[pscheduler.json_dump(limits_passed), task_uuid])
# Enable the task so the scheduler will schedule it.
try:
dbcursor_query("SELECT api_task_enable(%s)", [task_uuid])
except Exception:
log.exception()
return error("Failed to enable task %s. See system logs." %
task_uuid)
log.debug("Task enabled for scheduling.")
task_url = "%s/%s" % (request.base_url, task_uuid)
# Non-expanded gets just the URL
if not arg_boolean("expanded"):
return ok_json(task_url)
# Expanded gets a redirect to GET+expanded
params = []
for arg in ["detail", "pretty"]:
if arg_boolean(arg):
params.append(arg)
if params:
task_url += "?%s" % ("&".join(params))
return see_other(task_url)
else:
return not_allowed()
def schedule():
try:
range_start = arg_datetime('start')
range_end = arg_datetime('end')
except ValueError:
return bad_request('Invalid start or end time')
try:
task = arg_uuid("task")
except ValueError:
return bad_request('Invalid task UUID')
query = [
"""
SELECT
lower(times),
upper(times),
task,
run,
state_enum,
state_display,
task_json,
task_cli,
test_json,
tool_json,
errors
FROM schedule
WHERE times && tstzrange(%s, %s, '[)')
"""
]
args = [range_start, range_end]
if task is not None:
query.append("AND task = %s")
args.append(task)
try:
cursor = dbcursor_query(" ".join(query), args)
except Exception as ex:
log.exception()
return error(str(ex))
result = []
base_url = pscheduler.api_url_hostport(server_netloc(), "tasks/")
for row in cursor:
task_href = base_url + row[2]
run_href = "%s/runs/%s" % (task_href, row[3])
run = {
"start-time": pscheduler.datetime_as_iso8601(row[0]),
"end-time": pscheduler.datetime_as_iso8601(row[1]),
"href": run_href,
"result-href": "%s/result" % run_href,
"state": row[4],
"state-display": row[5],
"task": row[6],
"cli": row[7],
"test": row[8],
"tool": row[9],
"errors": row[10]
}
run["task"]["href"] = task_href
result.append(run)
return ok_json(result)