def __init__(
self,
transform # JSON JQTransformSpecification, assumed validated
):
# We modify this.
transform = copy.deepcopy(transform)
if isinstance(transform["script"], list):
script = "\n".join(transform["script"])
else:
script = transform["script"]
# Insert our function definition(s)
script_lines = [
"def classifiers:",
" ." + self.PRIVATE_KEY + ".classifiers",
";",
"def classifiers_has($value):",
" ." + self.PRIVATE_KEY + ".classifiers",
" | contains([$value])",
";",
"def default: %d;" % (self.DEFAULT_PRIORITY),
# This will be embedded in the task.
"def requested:",
" .priority",
";",
"def note($message):",
" if $message != null",
" then",
" ." + self.PRIVATE_KEY + ".diags += "
" [ \"\\($message | tostring)\" ]"
" else",
" .",
" end",
";",
"def priority:",
" ." + self.PRIVATE_KEY + ".priority",
";",
"def set($value; $message):",
" # TODO: Must be an integer",
" ." + self.PRIVATE_KEY + ".priority = $value",
" | note(\"\\($message) (Set to \\($value))\")",
";",
"def adjust($value; $message):",
" # TODO: Must be an integer",
" ." + self.PRIVATE_KEY + ".priority += $value",
" | note(\"\\($message) (\\(if $value > 0 then \"+\" else \"-\" end)\\($value | length))\")",
";",
script
]
transform["script"] = script_lines
self.transform = pscheduler.JQFilter(filter_spec=transform,
args=transform.get("args", {}),
groom=True)
def __init__(
self,
data # Data suitable for this class
):
valid, message = data_is_valid(data)
if not valid:
raise ValueError("Invalid data: %s" % message)
self.jqfilter = pscheduler.JQFilter(data["script"],
args=data.get("args", {}),
output_raw=data.get(
"output-raw", False))
def __init__(
self,
data # Data suitable for this class
):
valid, message = jq_data_is_valid(data)
if not valid:
raise ValueError("Invalid data: %s" % message)
self.jqfilter = pscheduler.JQFilter(
data["script"],
args=data.get("args", {})
# Don't bother with raw output. We don't care.
)
def __init__(
self,
transform # JSON JQTransformSpecification, assumed validated
):
# We modify this.
transform = copy.deepcopy(transform)
if isinstance(transform["script"], list):
script = "\n".join(transform["script"])
else:
script = transform["script"]
# One quirk of jq is that imports have to be at the top of the
# file. Pull any in the script out and move them to the top
# before we define our function or any user code.
import_include = r"((import|include) [^;]+;)"
# Put imports and includes at the top
lines = map(lambda x: x[0], re.findall(import_include, script))
# Insert our function definition(s)
lines.extend([
"def classifiers:",
" ." + self.PRIVATE_KEY + ".classifiers",
";",
"def change($message):",
" ." + self.PRIVATE_KEY + ".changed = true",
" | if $message != null",
" then",
" ." + self.PRIVATE_KEY + ".diags += [ $message | tostring ]",
" else",
" .",
" end",
";",
"def reject($message):",
" error(\"Task rejected: \" + ($message | tostring))",
";",
])
# Add the rest of the script without the imports and includes
lines.append(re.sub(import_include, "", script))
transform["script"] = lines
self.transform = pscheduler.JQFilter(
filter_spec=transform,
args=transform.get("args", {}),
)
def _jq_filter(transform):
if transform is None:
return None
script = transform["script"]
if isinstance(script, list):
script = "\n".join(script)
full_script = "def hint($name): ." + PRIVATE_KEY + ".hints[$name];" \
+ script
return pscheduler.JQFilter(full_script,
args=transform.get("args", {})
# Don't care about raw output. Not doing that.
)
def __init__(
self,
data # Data suitable for this class
):
valid, message = data_is_valid(data)
if not valid:
raise ValueError("Invalid data: %s" % message)
self.source = data['source']
self.bind = data.get('bind', None)
self.update = pscheduler.iso8601_as_timedelta(data['update'])
self.retry = pscheduler.iso8601_as_timedelta(data['retry'])
self.fail_state = data.get('fail-state', False)
try:
# This will raise a ValueError if it's wrong.
transform = data["transform"]
self.transform = pscheduler.JQFilter(transform["script"],
transform.get("args", {}),
output_raw=True)
except KeyError:
self.transform = None
self.exclusions = radix.Radix()
if 'exclude' in data:
try:
for excl in data['exclude']:
self.exclusions.add(excl)
except ValueError:
raise ValueError("Invalid IP or CIDR '%s'" % excl)
self.data_lock = threading.Lock()
self.updating = False
# TODO: Would be nice to support a timeout so the system
# doesn't sit for too long.
self.cidrs = radix.Radix()
self.length = 0
# Prime the timer with the epoch and do a first load of the list
self.next_attempt = datetime.datetime.utcfromtimestamp(0)
self.__populate_cidrs__()
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()
