def execute_rolling_restart(brokers,
jmxproxy_port,
jmxproxy_prefix,
check_interval,
check_count,
unhealthy_time_limit,
skip,
verbose,
pre_stop_task,
post_stop_task,
start_command,
stop_command,
ssh_password=None):
"""Execute the rolling restart on the specified brokers. It checks the
number of under replicated partitions on each broker, using JMXproxy.
The check is performed at constant intervals, and a broker will be restarted
when all the brokers are answering and are reporting zero under replicated
partitions.
:param brokers: the brokers that will be restarted
:type brokers: map of broker ids and host names
:param jmxproxy_port: HTTP port for JMXproxy
:type jmxproxy_port: integer
:param jmxproxy_prefix: HTTP prefix on the server for the JMXproxy queries
:type jmxproxy_prefix: string
:param check_interval: the number of seconds it will wait between each check
:type check_interval: integer
:param check_count: the number of times the check should be positive before
restarting the next broker
:type check_count: integer
:param unhealthy_time_limit: the maximum number of seconds it will wait for
the cluster to become stable before exiting with error
:type unhealthy_time_limit: integer
:param skip: the number of brokers to skip
:type skip: integer
:param verbose: print commend execution information
:type verbose: bool
:param pre_stop_task: a list of tasks to execute before running stop
:type pre_stop_task: list
:param post_stop_task: a list of task to execute after running stop
:type post_stop_task: list
:param start_command: the start command for kafka
:type start_command: string
:param stop_command: the stop command for kafka
:type stop_command: string
:param ssh_password: The ssh password to use if needed
:type ssh_password: string
"""
all_hosts = [b[1] for b in brokers]
for n, host in enumerate(all_hosts[skip:]):
with ssh(host=host,
forward_agent=True,
sudoable=True,
max_attempts=3,
max_timeout=2,
ssh_password=ssh_password) as connection:
execute_task(pre_stop_task, host)
wait_for_stable_cluster(
all_hosts,
jmxproxy_port,
jmxproxy_prefix,
check_interval,
1 if n == 0 else check_count,
unhealthy_time_limit,
)
print("Stopping {0} ({1}/{2})".format(host, n + 1,
len(all_hosts) - skip))
stop_broker(host, connection, stop_command, verbose)
execute_task(post_stop_task, host)
# we open a new SSH connection in case the hostname has a new IP
with ssh(host=host,
forward_agent=True,
sudoable=True,
max_attempts=3,
max_timeout=2,
ssh_password=ssh_password) as connection:
print("Starting {0} ({1}/{2})".format(host, n + 1,
len(all_hosts) - skip))
start_broker(host, connection, start_command, verbose)
# Wait before terminating the script
wait_for_stable_cluster(
all_hosts,
jmxproxy_port,
jmxproxy_prefix,
check_interval,
check_count,
unhealthy_time_limit,
)
def execute_rolling_restart(
brokers,
jolokia_port,
jolokia_prefix,
check_interval,
check_count,
unhealthy_time_limit,
skip,
verbose,
pre_stop_task,
post_stop_task,
start_command,
stop_command,
ssh_password=None
):
"""Execute the rolling restart on the specified brokers. It checks the
number of under replicated partitions on each broker, using Jolokia.
The check is performed at constant intervals, and a broker will be restarted
when all the brokers are answering and are reporting zero under replicated
partitions.
:param brokers: the brokers that will be restarted
:type brokers: map of broker ids and host names
:param jolokia_port: HTTP port for Jolokia
:type jolokia_port: integer
:param jolokia_prefix: HTTP prefix on the server for the Jolokia queries
:type jolokia_prefix: string
:param check_interval: the number of seconds it will wait between each check
:type check_interval: integer
:param check_count: the number of times the check should be positive before
restarting the next broker
:type check_count: integer
:param unhealthy_time_limit: the maximum number of seconds it will wait for
the cluster to become stable before exiting with error
:type unhealthy_time_limit: integer
:param skip: the number of brokers to skip
:type skip: integer
:param verbose: print commend execution information
:type verbose: bool
:param pre_stop_task: a list of tasks to execute before running stop
:type pre_stop_task: list
:param post_stop_task: a list of task to execute after running stop
:type post_stop_task: list
:param start_command: the start command for kafka
:type start_command: string
:param stop_command: the stop command for kafka
:type stop_command: string
:param ssh_password: The ssh password to use if needed
:type ssh_password: string
"""
all_hosts = [b[1] for b in brokers]
for n, host in enumerate(all_hosts[skip:]):
with ssh(host=host, forward_agent=True, sudoable=True, max_attempts=3, max_timeout=2,
ssh_password=ssh_password) as connection:
execute_task(pre_stop_task, host)
wait_for_stable_cluster(
all_hosts,
jolokia_port,
jolokia_prefix,
check_interval,
1 if n == 0 else check_count,
unhealthy_time_limit,
)
print("Stopping {0} ({1}/{2})".format(host, n + 1, len(all_hosts) - skip))
stop_broker(host, connection, stop_command, verbose)
execute_task(post_stop_task, host)
# we open a new SSH connection in case the hostname has a new IP
with ssh(host=host, forward_agent=True, sudoable=True, max_attempts=3, max_timeout=2,
ssh_password=ssh_password) as connection:
print("Starting {0} ({1}/{2})".format(host, n + 1, len(all_hosts) - skip))
start_broker(host, connection, start_command, verbose)
# Wait before terminating the script
wait_for_stable_cluster(
all_hosts,
jolokia_port,
jolokia_prefix,
check_interval,
check_count,
unhealthy_time_limit,
)