You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
502 lines
23 KiB
502 lines
23 KiB
# Licensed to the Apache Software Foundation (ASF) under one or more |
|
# contributor license agreements. See the NOTICE file distributed with |
|
# this work for additional information regarding copyright ownership. |
|
# The ASF licenses this file to You 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 json |
|
import os.path |
|
import random |
|
import signal |
|
import time |
|
|
|
import requests |
|
from ducktape.errors import DucktapeError |
|
from ducktape.services.service import Service |
|
from ducktape.utils.util import wait_until |
|
|
|
from kafkatest.directory_layout.kafka_path import KafkaPathResolverMixin |
|
|
|
|
|
class ConnectServiceBase(KafkaPathResolverMixin, Service): |
|
"""Base class for Kafka Connect services providing some common settings and functionality""" |
|
|
|
PERSISTENT_ROOT = "/mnt/connect" |
|
CONFIG_FILE = os.path.join(PERSISTENT_ROOT, "connect.properties") |
|
# The log file contains normal log4j logs written using a file appender. stdout and stderr are handled separately |
|
# so they can be used for other output, e.g. verifiable source & sink. |
|
LOG_FILE = os.path.join(PERSISTENT_ROOT, "connect.log") |
|
STDOUT_FILE = os.path.join(PERSISTENT_ROOT, "connect.stdout") |
|
STDERR_FILE = os.path.join(PERSISTENT_ROOT, "connect.stderr") |
|
LOG4J_CONFIG_FILE = os.path.join(PERSISTENT_ROOT, "connect-log4j.properties") |
|
PID_FILE = os.path.join(PERSISTENT_ROOT, "connect.pid") |
|
EXTERNAL_CONFIGS_FILE = os.path.join(PERSISTENT_ROOT, "connect-external-configs.properties") |
|
CONNECT_REST_PORT = 8083 |
|
HEAP_DUMP_FILE = os.path.join(PERSISTENT_ROOT, "connect_heap_dump.bin") |
|
|
|
# Currently the Connect worker supports waiting on three modes: |
|
STARTUP_MODE_INSTANT = 'INSTANT' |
|
"""STARTUP_MODE_INSTANT: Start Connect worker and return immediately""" |
|
STARTUP_MODE_LOAD = 'LOAD' |
|
"""STARTUP_MODE_LOAD: Start Connect worker and return after discovering and loading plugins""" |
|
STARTUP_MODE_LISTEN = 'LISTEN' |
|
"""STARTUP_MODE_LISTEN: Start Connect worker and return after opening the REST port.""" |
|
|
|
logs = { |
|
"connect_log": { |
|
"path": LOG_FILE, |
|
"collect_default": True}, |
|
"connect_stdout": { |
|
"path": STDOUT_FILE, |
|
"collect_default": False}, |
|
"connect_stderr": { |
|
"path": STDERR_FILE, |
|
"collect_default": True}, |
|
"connect_heap_dump_file": { |
|
"path": HEAP_DUMP_FILE, |
|
"collect_default": True} |
|
} |
|
|
|
def __init__(self, context, num_nodes, kafka, files, startup_timeout_sec = 60): |
|
super(ConnectServiceBase, self).__init__(context, num_nodes) |
|
self.kafka = kafka |
|
self.security_config = kafka.security_config.client_config() |
|
self.files = files |
|
self.startup_mode = self.STARTUP_MODE_LISTEN |
|
self.startup_timeout_sec = startup_timeout_sec |
|
self.environment = {} |
|
self.external_config_template_func = None |
|
|
|
def pids(self, node): |
|
"""Return process ids for Kafka Connect processes.""" |
|
try: |
|
return [pid for pid in node.account.ssh_capture("cat " + self.PID_FILE, callback=int)] |
|
except: |
|
return [] |
|
|
|
def set_configs(self, config_template_func, connector_config_templates=None): |
|
""" |
|
Set configurations for the worker and the connector to run on |
|
it. These are not provided in the constructor because the worker |
|
config generally needs access to ZK/Kafka services to |
|
create the configuration. |
|
""" |
|
self.config_template_func = config_template_func |
|
self.connector_config_templates = connector_config_templates |
|
|
|
def set_external_configs(self, external_config_template_func): |
|
""" |
|
Set the properties that will be written in the external file properties |
|
as used by the org.apache.kafka.common.config.provider.FileConfigProvider. |
|
When this is used, the worker configuration must also enable the FileConfigProvider. |
|
This is not provided in the constructor in case the worker |
|
config generally needs access to ZK/Kafka services to |
|
create the configuration. |
|
""" |
|
self.external_config_template_func = external_config_template_func |
|
|
|
def listening(self, node): |
|
try: |
|
self.list_connectors(node) |
|
self.logger.debug("Connect worker started serving REST at: '%s:%s')", node.account.hostname, |
|
self.CONNECT_REST_PORT) |
|
return True |
|
except requests.exceptions.ConnectionError: |
|
self.logger.debug("REST resources are not loaded yet") |
|
return False |
|
|
|
def start(self, mode=STARTUP_MODE_LISTEN): |
|
self.startup_mode = mode |
|
super(ConnectServiceBase, self).start() |
|
|
|
def start_and_return_immediately(self, node, worker_type, remote_connector_configs): |
|
cmd = self.start_cmd(node, remote_connector_configs) |
|
self.logger.debug("Connect %s command: %s", worker_type, cmd) |
|
node.account.ssh(cmd) |
|
|
|
def start_and_wait_to_load_plugins(self, node, worker_type, remote_connector_configs): |
|
with node.account.monitor_log(self.LOG_FILE) as monitor: |
|
self.start_and_return_immediately(node, worker_type, remote_connector_configs) |
|
monitor.wait_until('Kafka version', timeout_sec=self.startup_timeout_sec, |
|
err_msg="Never saw message indicating Kafka Connect finished startup on node: " + |
|
"%s in condition mode: %s" % (str(node.account), self.startup_mode)) |
|
|
|
def start_and_wait_to_start_listening(self, node, worker_type, remote_connector_configs): |
|
self.start_and_return_immediately(node, worker_type, remote_connector_configs) |
|
wait_until(lambda: self.listening(node), timeout_sec=self.startup_timeout_sec, |
|
err_msg="Kafka Connect failed to start on node: %s in condition mode: %s" % |
|
(str(node.account), self.startup_mode)) |
|
|
|
def stop_node(self, node, clean_shutdown=True): |
|
self.logger.info((clean_shutdown and "Cleanly" or "Forcibly") + " stopping Kafka Connect on " + str(node.account)) |
|
pids = self.pids(node) |
|
sig = signal.SIGTERM if clean_shutdown else signal.SIGKILL |
|
|
|
for pid in pids: |
|
node.account.signal(pid, sig, allow_fail=True) |
|
if clean_shutdown: |
|
for pid in pids: |
|
wait_until(lambda: not node.account.alive(pid), timeout_sec=self.startup_timeout_sec, err_msg="Kafka Connect process on " + str( |
|
node.account) + " took too long to exit") |
|
|
|
node.account.ssh("rm -f " + self.PID_FILE, allow_fail=False) |
|
|
|
def restart(self, clean_shutdown=True): |
|
# We don't want to do any clean up here, just restart the process. |
|
for node in self.nodes: |
|
self.logger.info("Restarting Kafka Connect on " + str(node.account)) |
|
self.restart_node(node, clean_shutdown) |
|
|
|
def restart_node(self, node, clean_shutdown=True): |
|
self.stop_node(node, clean_shutdown) |
|
self.start_node(node) |
|
|
|
def clean_node(self, node): |
|
node.account.kill_process("connect", clean_shutdown=False, allow_fail=True) |
|
self.security_config.clean_node(node) |
|
other_files = " ".join(self.config_filenames() + self.files) |
|
node.account.ssh("rm -rf -- %s %s" % (ConnectServiceBase.PERSISTENT_ROOT, other_files), allow_fail=False) |
|
|
|
def config_filenames(self): |
|
return [os.path.join(self.PERSISTENT_ROOT, "connect-connector-" + str(idx) + ".properties") for idx, template in enumerate(self.connector_config_templates or [])] |
|
|
|
def list_connectors(self, node=None, **kwargs): |
|
return self._rest_with_retry('/connectors', node=node, **kwargs) |
|
|
|
def create_connector(self, config, node=None, **kwargs): |
|
create_request = { |
|
'name': config['name'], |
|
'config': config |
|
} |
|
return self._rest_with_retry('/connectors', create_request, node=node, method="POST", **kwargs) |
|
|
|
def get_connector(self, name, node=None, **kwargs): |
|
return self._rest_with_retry('/connectors/' + name, node=node, **kwargs) |
|
|
|
def get_connector_config(self, name, node=None, **kwargs): |
|
return self._rest_with_retry('/connectors/' + name + '/config', node=node, **kwargs) |
|
|
|
def set_connector_config(self, name, config, node=None, **kwargs): |
|
# Unlike many other calls, a 409 when setting a connector config is expected if the connector already exists. |
|
# However, we also might see 409s for other reasons (e.g. rebalancing). So we still perform retries at the cost |
|
# of tests possibly taking longer to ultimately fail. Tests that care about this can explicitly override the |
|
# number of retries. |
|
return self._rest_with_retry('/connectors/' + name + '/config', config, node=node, method="PUT", **kwargs) |
|
|
|
def get_connector_tasks(self, name, node=None, **kwargs): |
|
return self._rest_with_retry('/connectors/' + name + '/tasks', node=node, **kwargs) |
|
|
|
def delete_connector(self, name, node=None, **kwargs): |
|
return self._rest_with_retry('/connectors/' + name, node=node, method="DELETE", **kwargs) |
|
|
|
def get_connector_status(self, name, node=None): |
|
return self._rest('/connectors/' + name + '/status', node=node) |
|
|
|
def restart_connector(self, name, node=None, **kwargs): |
|
return self._rest_with_retry('/connectors/' + name + '/restart', node=node, method="POST", **kwargs) |
|
|
|
def restart_task(self, connector_name, task_id, node=None): |
|
return self._rest('/connectors/' + connector_name + '/tasks/' + str(task_id) + '/restart', node=node, method="POST") |
|
|
|
def pause_connector(self, name, node=None): |
|
return self._rest('/connectors/' + name + '/pause', node=node, method="PUT") |
|
|
|
def resume_connector(self, name, node=None): |
|
return self._rest('/connectors/' + name + '/resume', node=node, method="PUT") |
|
|
|
def list_connector_plugins(self, node=None): |
|
return self._rest('/connector-plugins/', node=node) |
|
|
|
def validate_config(self, connector_type, validate_request, node=None): |
|
return self._rest('/connector-plugins/' + connector_type + '/config/validate', validate_request, node=node, method="PUT") |
|
|
|
def _rest(self, path, body=None, node=None, method="GET"): |
|
if node is None: |
|
node = random.choice(self.nodes) |
|
|
|
meth = getattr(requests, method.lower()) |
|
url = self._base_url(node) + path |
|
self.logger.debug("Kafka Connect REST request: %s %s %s %s", node.account.hostname, url, method, body) |
|
resp = meth(url, json=body) |
|
self.logger.debug("%s %s response: %d", url, method, resp.status_code) |
|
if resp.status_code > 400: |
|
self.logger.debug("Connect REST API error for %s: %d %s", resp.url, resp.status_code, resp.text) |
|
raise ConnectRestError(resp.status_code, resp.text, resp.url) |
|
if resp.status_code == 204 or resp.status_code == 202: |
|
return None |
|
else: |
|
return resp.json() |
|
|
|
def _rest_with_retry(self, path, body=None, node=None, method="GET", retries=40, retry_backoff=.25): |
|
""" |
|
Invokes a REST API with retries for errors that may occur during normal operation (notably 409 CONFLICT |
|
responses that can occur due to rebalancing or 404 when the connect resources are not initialized yet). |
|
""" |
|
exception_to_throw = None |
|
for i in range(0, retries + 1): |
|
try: |
|
return self._rest(path, body, node, method) |
|
except ConnectRestError as e: |
|
exception_to_throw = e |
|
if e.status != 409 and e.status != 404: |
|
break |
|
time.sleep(retry_backoff) |
|
raise exception_to_throw |
|
|
|
def _base_url(self, node): |
|
return 'http://' + node.account.externally_routable_ip + ':' + str(self.CONNECT_REST_PORT) |
|
|
|
def append_to_environment_variable(self, envvar, value): |
|
env_opts = self.environment[envvar] |
|
if env_opts is None: |
|
env_opts = "\"%s\"" % value |
|
else: |
|
env_opts = "\"%s %s\"" % (env_opts.strip('\"'), value) |
|
self.environment[envvar] = env_opts |
|
|
|
|
|
class ConnectStandaloneService(ConnectServiceBase): |
|
"""Runs Kafka Connect in standalone mode.""" |
|
|
|
def __init__(self, context, kafka, files, startup_timeout_sec = 60): |
|
super(ConnectStandaloneService, self).__init__(context, 1, kafka, files, startup_timeout_sec) |
|
|
|
# For convenience since this service only makes sense with a single node |
|
@property |
|
def node(self): |
|
return self.nodes[0] |
|
|
|
def start_cmd(self, node, connector_configs): |
|
cmd = "( export KAFKA_LOG4J_OPTS=\"-Dlog4j.configuration=file:%s\"; " % self.LOG4J_CONFIG_FILE |
|
heap_kafka_opts = "-XX:+HeapDumpOnOutOfMemoryError -XX:HeapDumpPath=%s" % \ |
|
self.logs["connect_heap_dump_file"]["path"] |
|
other_kafka_opts = self.security_config.kafka_opts.strip('\"') |
|
cmd += "export KAFKA_OPTS=\"%s %s\"; " % (heap_kafka_opts, other_kafka_opts) |
|
for envvar in self.environment: |
|
cmd += "export %s=%s; " % (envvar, str(self.environment[envvar])) |
|
cmd += "%s %s " % (self.path.script("connect-standalone.sh", node), self.CONFIG_FILE) |
|
cmd += " ".join(connector_configs) |
|
cmd += " & echo $! >&3 ) 1>> %s 2>> %s 3> %s" % (self.STDOUT_FILE, self.STDERR_FILE, self.PID_FILE) |
|
return cmd |
|
|
|
def start_node(self, node): |
|
node.account.ssh("mkdir -p %s" % self.PERSISTENT_ROOT, allow_fail=False) |
|
|
|
self.security_config.setup_node(node) |
|
if self.external_config_template_func: |
|
node.account.create_file(self.EXTERNAL_CONFIGS_FILE, self.external_config_template_func(node)) |
|
node.account.create_file(self.CONFIG_FILE, self.config_template_func(node)) |
|
node.account.create_file(self.LOG4J_CONFIG_FILE, self.render('connect_log4j.properties', log_file=self.LOG_FILE)) |
|
remote_connector_configs = [] |
|
for idx, template in enumerate(self.connector_config_templates): |
|
target_file = os.path.join(self.PERSISTENT_ROOT, "connect-connector-" + str(idx) + ".properties") |
|
node.account.create_file(target_file, template) |
|
remote_connector_configs.append(target_file) |
|
|
|
self.logger.info("Starting Kafka Connect standalone process on " + str(node.account)) |
|
if self.startup_mode == self.STARTUP_MODE_LOAD: |
|
self.start_and_wait_to_load_plugins(node, 'standalone', remote_connector_configs) |
|
elif self.startup_mode == self.STARTUP_MODE_INSTANT: |
|
self.start_and_return_immediately(node, 'standalone', remote_connector_configs) |
|
else: |
|
# The default mode is to wait until the complete startup of the worker |
|
self.start_and_wait_to_start_listening(node, 'standalone', remote_connector_configs) |
|
|
|
if len(self.pids(node)) == 0: |
|
raise RuntimeError("No process ids recorded") |
|
|
|
|
|
class ConnectDistributedService(ConnectServiceBase): |
|
"""Runs Kafka Connect in distributed mode.""" |
|
|
|
def __init__(self, context, num_nodes, kafka, files, offsets_topic="connect-offsets", |
|
configs_topic="connect-configs", status_topic="connect-status", startup_timeout_sec = 60): |
|
super(ConnectDistributedService, self).__init__(context, num_nodes, kafka, files, startup_timeout_sec) |
|
self.offsets_topic = offsets_topic |
|
self.configs_topic = configs_topic |
|
self.status_topic = status_topic |
|
|
|
# connector_configs argument is intentionally ignored in distributed service. |
|
def start_cmd(self, node, connector_configs): |
|
cmd = "( export KAFKA_LOG4J_OPTS=\"-Dlog4j.configuration=file:%s\"; " % self.LOG4J_CONFIG_FILE |
|
heap_kafka_opts = "-XX:+HeapDumpOnOutOfMemoryError -XX:HeapDumpPath=%s" % \ |
|
self.logs["connect_heap_dump_file"]["path"] |
|
other_kafka_opts = self.security_config.kafka_opts.strip('\"') |
|
cmd += "export KAFKA_OPTS=\"%s %s\"; " % (heap_kafka_opts, other_kafka_opts) |
|
for envvar in self.environment: |
|
cmd += "export %s=%s; " % (envvar, str(self.environment[envvar])) |
|
cmd += "%s %s " % (self.path.script("connect-distributed.sh", node), self.CONFIG_FILE) |
|
cmd += " & echo $! >&3 ) 1>> %s 2>> %s 3> %s" % (self.STDOUT_FILE, self.STDERR_FILE, self.PID_FILE) |
|
return cmd |
|
|
|
def start_node(self, node): |
|
node.account.ssh("mkdir -p %s" % self.PERSISTENT_ROOT, allow_fail=False) |
|
|
|
self.security_config.setup_node(node) |
|
if self.external_config_template_func: |
|
node.account.create_file(self.EXTERNAL_CONFIGS_FILE, self.external_config_template_func(node)) |
|
node.account.create_file(self.CONFIG_FILE, self.config_template_func(node)) |
|
node.account.create_file(self.LOG4J_CONFIG_FILE, self.render('connect_log4j.properties', log_file=self.LOG_FILE)) |
|
if self.connector_config_templates: |
|
raise DucktapeError("Config files are not valid in distributed mode, submit connectors via the REST API") |
|
|
|
self.logger.info("Starting Kafka Connect distributed process on " + str(node.account)) |
|
if self.startup_mode == self.STARTUP_MODE_LOAD: |
|
self.start_and_wait_to_load_plugins(node, 'distributed', '') |
|
elif self.startup_mode == self.STARTUP_MODE_INSTANT: |
|
self.start_and_return_immediately(node, 'distributed', '') |
|
else: |
|
# The default mode is to wait until the complete startup of the worker |
|
self.start_and_wait_to_start_listening(node, 'distributed', '') |
|
|
|
if len(self.pids(node)) == 0: |
|
raise RuntimeError("No process ids recorded") |
|
|
|
|
|
class ErrorTolerance(object): |
|
ALL = "all" |
|
NONE = "none" |
|
|
|
|
|
class ConnectRestError(RuntimeError): |
|
def __init__(self, status, msg, url): |
|
self.status = status |
|
self.message = msg |
|
self.url = url |
|
|
|
def __unicode__(self): |
|
return "Kafka Connect REST call failed: returned " + self.status + " for " + self.url + ". Response: " + self.message |
|
|
|
|
|
class VerifiableConnector(object): |
|
def messages(self): |
|
""" |
|
Collect and parse the logs from Kafka Connect nodes. Return a list containing all parsed JSON messages generated by |
|
this source. |
|
""" |
|
self.logger.info("Collecting messages from log of %s %s", type(self).__name__, self.name) |
|
records = [] |
|
for node in self.cc.nodes: |
|
for line in node.account.ssh_capture('cat ' + self.cc.STDOUT_FILE): |
|
try: |
|
data = json.loads(line) |
|
except ValueError: |
|
self.logger.debug("Ignoring unparseable line: %s", line) |
|
continue |
|
# Filter to only ones matching our name to support multiple verifiable producers |
|
if data['name'] != self.name: |
|
continue |
|
data['node'] = node |
|
records.append(data) |
|
return records |
|
|
|
def stop(self): |
|
self.logger.info("Destroying connector %s %s", type(self).__name__, self.name) |
|
self.cc.delete_connector(self.name) |
|
|
|
|
|
class VerifiableSource(VerifiableConnector): |
|
""" |
|
Helper class for running a verifiable source connector on a Kafka Connect cluster and analyzing the output. |
|
""" |
|
|
|
def __init__(self, cc, name="verifiable-source", tasks=1, topic="verifiable", throughput=1000): |
|
self.cc = cc |
|
self.logger = self.cc.logger |
|
self.name = name |
|
self.tasks = tasks |
|
self.topic = topic |
|
self.throughput = throughput |
|
|
|
def committed_messages(self): |
|
return filter(lambda m: 'committed' in m and m['committed'], self.messages()) |
|
|
|
def sent_messages(self): |
|
return filter(lambda m: 'committed' not in m or not m['committed'], self.messages()) |
|
|
|
def start(self): |
|
self.logger.info("Creating connector VerifiableSourceConnector %s", self.name) |
|
self.cc.create_connector({ |
|
'name': self.name, |
|
'connector.class': 'org.apache.kafka.connect.tools.VerifiableSourceConnector', |
|
'tasks.max': self.tasks, |
|
'topic': self.topic, |
|
'throughput': self.throughput |
|
}) |
|
|
|
|
|
class VerifiableSink(VerifiableConnector): |
|
""" |
|
Helper class for running a verifiable sink connector on a Kafka Connect cluster and analyzing the output. |
|
""" |
|
|
|
def __init__(self, cc, name="verifiable-sink", tasks=1, topics=["verifiable"]): |
|
self.cc = cc |
|
self.logger = self.cc.logger |
|
self.name = name |
|
self.tasks = tasks |
|
self.topics = topics |
|
|
|
def flushed_messages(self): |
|
return filter(lambda m: 'flushed' in m and m['flushed'], self.messages()) |
|
|
|
def received_messages(self): |
|
return filter(lambda m: 'flushed' not in m or not m['flushed'], self.messages()) |
|
|
|
def start(self): |
|
self.logger.info("Creating connector VerifiableSinkConnector %s", self.name) |
|
self.cc.create_connector({ |
|
'name': self.name, |
|
'connector.class': 'org.apache.kafka.connect.tools.VerifiableSinkConnector', |
|
'tasks.max': self.tasks, |
|
'topics': ",".join(self.topics) |
|
}) |
|
|
|
class MockSink(object): |
|
|
|
def __init__(self, cc, topics, mode=None, delay_sec=10, name="mock-sink"): |
|
self.cc = cc |
|
self.logger = self.cc.logger |
|
self.name = name |
|
self.mode = mode |
|
self.delay_sec = delay_sec |
|
self.topics = topics |
|
|
|
def start(self): |
|
self.logger.info("Creating connector MockSinkConnector %s", self.name) |
|
self.cc.create_connector({ |
|
'name': self.name, |
|
'connector.class': 'org.apache.kafka.connect.tools.MockSinkConnector', |
|
'tasks.max': 1, |
|
'topics': ",".join(self.topics), |
|
'mock_mode': self.mode, |
|
'delay_ms': self.delay_sec * 1000 |
|
}) |
|
|
|
class MockSource(object): |
|
|
|
def __init__(self, cc, mode=None, delay_sec=10, name="mock-source"): |
|
self.cc = cc |
|
self.logger = self.cc.logger |
|
self.name = name |
|
self.mode = mode |
|
self.delay_sec = delay_sec |
|
|
|
def start(self): |
|
self.logger.info("Creating connector MockSourceConnector %s", self.name) |
|
self.cc.create_connector({ |
|
'name': self.name, |
|
'connector.class': 'org.apache.kafka.connect.tools.MockSourceConnector', |
|
'tasks.max': 1, |
|
'mock_mode': self.mode, |
|
'delay_ms': self.delay_sec * 1000 |
|
})
|
|
|