zuul/zuul/executor/client.py
James E. Blair 1f026bd49c Finish circular dependency refactor
This change completes the circular dependency refactor.

The principal change is that queue items may now include
more than one change simultaneously in the case of circular
dependencies.

In dependent pipelines, the two-phase reporting process is
simplified because it happens during processing of a single
item.

In independent pipelines, non-live items are still used for
linear depnedencies, but multi-change items are used for
circular dependencies.

Previously changes were enqueued recursively and then
bundles were made out of the resulting items.  Since we now
need to enqueue entire cycles in one queue item, the
dependency graph generation is performed at the start of
enqueing the first change in a cycle.

Some tests exercise situations where Zuul is processing
events for old patchsets of changes.  The new change query
sequence mentioned in the previous paragraph necessitates
more accurate information about out-of-date patchsets than
the previous sequence, therefore the Gerrit driver has been
updated to query and return more data about non-current
patchsets.

This change is not backwards compatible with the existing
ZK schema, and will require Zuul systems delete all pipeline
states during the upgrade.  A later change will implement
a helper command for this.

All backwards compatability handling for the last several
model_api versions which were added to prepare for this
upgrade have been removed.  In general, all model data
structures involving frozen jobs are now indexed by the
frozen job's uuid and no longer include the job name since
a job name no longer uniquely identifies a job in a buildset
(either the uuid or the (job name, change) tuple must be
used to identify it).

Job deduplication is simplified and now only needs to
consider jobs within the same buildset.

The fake github driver had a bug (fakegithub.py line 694) where
it did not correctly increment the check run counter, so our
tests that verified that we closed out obsolete check runs
when re-enqueing were not valid.  This has been corrected, and
in doing so, has necessitated some changes around quiet dequeing
when we re-enqueue a change.

The reporting in several drivers has been updated to support
reporting information about multiple changes in a queue item.

Change-Id: I0b9e4d3f9936b1e66a08142fc36866269dc287f1
Depends-On: https://review.opendev.org/907627
2024-02-09 07:39:40 -08:00

318 lines
13 KiB
Python

# Copyright 2012 Hewlett-Packard Development Company, L.P.
#
# Licensed under the Apache License, Version 2.0 (the "License"); you may
# not use this file except in compliance with the License. You may obtain
# a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
# License for the specific language governing permissions and limitations
# under the License.
import logging
import time
from uuid import uuid4
import zuul.executor.common
from zuul.lib.logutil import get_annotated_logger
from zuul.model import (
Build,
BuildCompletedEvent,
BuildRequest,
BuildStartedEvent,
PRIORITY_MAP,
)
from zuul.zk.event_queues import PipelineResultEventQueue
from zuul.zk.executor import ExecutorApi
from zuul.zk.exceptions import JobRequestNotFound
import zuul.lib.tracing as tracing
from kazoo.exceptions import BadVersionError
from opentelemetry import trace
class ExecutorClient(object):
log = logging.getLogger("zuul.ExecutorClient")
_executor_api_class = ExecutorApi
def __init__(self, config, sched):
self.config = config
self.sched = sched
self.executor_api = self._executor_api_class(self.sched.zk_client)
self.result_events = PipelineResultEventQueue.createRegistry(
self.sched.zk_client
)
def stop(self):
self.log.debug("Stopping")
def execute(self, job, nodes, item, pipeline, executor_zone,
dependent_changes=[], merger_items=[]):
log = get_annotated_logger(self.log, item.event)
tracer = trace.get_tracer("zuul")
uuid = str(uuid4().hex)
log.info(
"Execute job %s (uuid: %s) on nodes %s for %s "
"with dependent changes %s",
job, uuid, nodes, item, dependent_changes)
params = zuul.executor.common.construct_build_params(
uuid, self.sched.connections,
job, item, pipeline, dependent_changes, merger_items,
redact_secrets_and_keys=False)
# TODO: deprecate and remove this variable?
params["zuul"]["_inheritance_path"] = list(job.inheritance_path)
semaphore_handler = item.pipeline.tenant.semaphore_handler
params['semaphore_handle'] = semaphore_handler.getSemaphoreHandle(
item, job)
parent_span = tracing.restoreSpan(item.current_build_set.span_info)
execute_time = time.time()
with trace.use_span(parent_span):
build_span = tracer.start_span("Build", start_time=execute_time)
build_span_info = tracing.getSpanInfo(build_span)
build = Build.new(
pipeline.manager.current_context,
job=job,
build_set=item.current_build_set,
uuid=uuid,
execute_time=execute_time,
span_info=build_span_info,
zuul_event_id=item.event.zuul_event_id,
)
log.debug("Adding build %s of job %s to item %s",
build, job, item)
item.addBuild(job, build)
if job.name == 'noop':
data = {"start_time": time.time()}
started_event = BuildStartedEvent(
build.uuid, build.build_set.uuid, job.uuid,
None, data, zuul_event_id=build.zuul_event_id)
self.result_events[pipeline.tenant.name][pipeline.name].put(
started_event
)
result = {"result": "SUCCESS", "end_time": time.time()}
completed_event = BuildCompletedEvent(
build.uuid, build.build_set.uuid, job.uuid,
None, result, zuul_event_id=build.zuul_event_id)
self.result_events[pipeline.tenant.name][pipeline.name].put(
completed_event
)
return
# Update zuul attempts after addBuild above to ensure build_set
# is up to date.
attempts = build.build_set.getTries(job)
params["zuul"]['attempts'] = attempts
params['zuul']['max_attempts'] = job.attempts
# TODO (swestphahl): Remove deprecated 'max_attempts' parameter
params['max_attempts'] = job.attempts
# Store the NodeRequest ID in the job arguments, so we can look it up
# on the executor side to lock the nodes.
req_id = build.build_set.getJobNodeRequestID(job)
if isinstance(req_id, dict):
# This is a stop-gap. It is possible for this to happen
# if a queue item completes all its builds and is removed
# while one of its builds is deduplicated in another queue
# item in an independent pipeline. The bundle refactor
# work will remove this possibility at which point this
# code can be removed. In the mean time, if we encounter
# this case, restart the build to try to keep things
# moving.
self.log.error(
"Attempt to start build with deduplicated node request ID "
f"{req_id}")
data = {"start_time": time.time()}
started_event = BuildStartedEvent(
build.uuid, build.build_set.uuid, job.uuid,
None, data, zuul_event_id=build.zuul_event_id)
self.result_events[pipeline.tenant.name][pipeline.name].put(
started_event
)
result = {"result": None, "end_time": time.time()}
completed_event = BuildCompletedEvent(
build.uuid, build.build_set.uuid, job.uuid,
None, result, zuul_event_id=build.zuul_event_id)
self.result_events[pipeline.tenant.name][pipeline.name].put(
completed_event
)
return
if req_id:
params["noderequest_id"] = req_id
zone_known = False
if executor_zone:
# Check the component registry for executors subscribed to this
# zone
for comp in self.sched.component_registry.all(kind="executor"):
if comp.zone == executor_zone:
zone_known = True
break
if not zone_known:
self.log.warning(
"Job requested '%s' zuul-executor zone, but no "
"zuul-executors found for this zone; ignoring zone "
"request", executor_zone)
# Fall back to the default zone
executor_zone = None
with trace.use_span(build_span):
request = BuildRequest(
uuid=uuid,
build_set_uuid=build.build_set.uuid,
job_uuid=job.uuid,
tenant_name=build.build_set.item.pipeline.tenant.name,
pipeline_name=build.build_set.item.pipeline.name,
zone=executor_zone,
event_id=item.event.zuul_event_id,
precedence=PRIORITY_MAP[pipeline.precedence],
)
self.executor_api.submit(request, params)
build.updateAttributes(pipeline.manager.current_context,
build_request_ref=request.path)
def cancel(self, build):
log = get_annotated_logger(self.log, build.zuul_event_id,
build=build.uuid)
# Returns whether a running build was canceled
log.info("Cancel build %s for job %s", build, build.job)
build.updateAttributes(
build.build_set.item.pipeline.manager.current_context,
canceled=True)
if not build.build_request_ref:
log.debug("Build has not been submitted to ZooKeeper")
return False
build_request = self.executor_api.get(build.build_request_ref)
if build_request:
log.debug("Canceling build request %s", build_request)
# If we can acquire the build request lock here, the build wasn't
# picked up by any executor server yet. With acquiring the lock
# we prevent the executor server from picking up the build so we
# can cancel it before it will run.
if self.executor_api.lock(build_request, blocking=False):
log.debug(
"Canceling build %s directly because it is not locked by "
"any executor",
build_request,
)
# Mark the build request as complete and forward the event to
# the scheduler, so the executor server doesn't pick up the
# request. The build will be deleted from the scheduler when it
# picks up the BuildCompletedEvent.
try:
build_request.state = BuildRequest.COMPLETED
self.executor_api.update(build_request)
result = {"result": "CANCELED", "end_time": time.time()}
tenant_name = build.build_set.item.pipeline.tenant.name
pipeline_name = build.build_set.item.pipeline.name
event = BuildCompletedEvent(
build_request.uuid, build_request.build_set_uuid,
build_request.job_uuid,
build_request.path, result)
self.result_events[tenant_name][pipeline_name].put(event)
finally:
self.executor_api.unlock(build_request)
else:
log.debug(
"Sending cancel request for build %s because it is locked",
build_request,
)
# If the build request is locked, schedule a cancel request in
# the executor server.
self.executor_api.requestCancel(build_request)
log.debug("Canceled build")
return True
return False
def resumeBuild(self, build):
log = get_annotated_logger(self.log, build.zuul_event_id)
if not build.build_request_ref:
log.debug("Build has not been submitted")
return False
build_request = self.executor_api.get(build.build_request_ref)
if build_request:
log.debug("Requesting resume for build %s", build)
self.executor_api.requestResume(build_request)
return True
return False
def removeBuild(self, build):
log = get_annotated_logger(self.log, build.zuul_event_id)
log.debug("Removing build %s", build.uuid)
if not build.build_request_ref:
log.debug("Build %s has not been submitted to ZooKeeper",
build.uuid)
return
build_request = self.executor_api.get(build.build_request_ref)
if build_request:
# TODO (felix): We could directly remove the build request via
# its path in ZK to spare a read operation. Usually there should
# be no need to look up the build request object from ZooKeeper
# just to immediately remove it.
self.executor_api.remove(build_request)
def cleanupLostBuildRequests(self):
for build_request in self.executor_api.lostRequests():
try:
self.cleanupLostBuildRequest(build_request)
except Exception:
self.log.exception("Exception cleaning up lost build request:")
def cleanupLostBuildRequest(self, build_request):
result = {"result": "ABORTED"}
# TODO (felix): Once the builds are stored in ZooKeeper, we can store
# the end_time directly on the build. But for now we have to use the
# result dict for that.
result["end_time"] = time.time()
build_request.state = BuildRequest.COMPLETED
try:
self.executor_api.update(build_request)
except JobRequestNotFound as e:
self.log.warning("Could not complete build: %s", str(e))
# In case we re-created the lock directory, still remove
# the request for the side effect of removing the lock.
self.executor_api.remove(build_request)
return
except BadVersionError:
# There could be a race condition:
# The build is found by lost_builds in state RUNNING
# but gets completed/unlocked before the is_locked()
# check. Since we use the znode version, the update
# will fail in this case and we can simply ignore the
# exception.
return
# No need to unlock the build, as it is by definition unlocked
event = BuildCompletedEvent(
build_request.uuid, build_request.build_set_uuid,
build_request.job_uuid,
build_request.path, result)
self.result_events[build_request.tenant_name][
build_request.pipeline_name].put(event)