def _check_sla(self, hostnames, grouping_function, percentage, duration):
"""Check if the provided list of hosts passes the job uptime SLA check.
This is an all-or-nothing check, meaning that all provided hosts must pass their job
SLA check for the maintenance to proceed.
:param hostnames: list of host names to check SLA for
:type hostnames: list of strings
:param grouping_function: grouping function to apply to the given hosts
:type grouping_function: function
:param percentage: SLA uptime percentage override
:type percentage: float
:param duration: SLA uptime duration override
:type duration: twitter.common.quantity.Amount
:rtype: set of unsafe hosts
"""
vector = self._client.sla_get_safe_domain_vector(self.SLA_MIN_JOB_INSTANCE_COUNT, hostnames)
host_groups = vector.probe_hosts(percentage, duration.as_(Time.SECONDS), grouping_function)
unsafe_hostnames = set()
# Given that maintenance is performed 1 group at a time, any result longer than 1 group
# should be considered a batch failure.
if host_groups:
if len(host_groups) > 1:
log.error("Illegal multiple groups detected in SLA results. Skipping hosts: %s" % hostnames)
return set(hostnames)
results, unsafe_hostnames = format_sla_results(host_groups, unsafe_only=True)
if results:
print_results(results)
return unsafe_hostnames
return unsafe_hostnames
def _check_sla(self, hostnames, grouping_function, percentage, duration):
"""Check if the provided list of hosts passes the job uptime SLA check.
This is an all-or-nothing check, meaning that all provided hosts must pass their job
SLA check for the maintenance to proceed.
:param hostnames: list of host names to check SLA for
:type hostnames: list of strings
:param grouping_function: grouping function to apply to the given hosts
:type grouping_function: function
:param percentage: SLA uptime percentage override
:type percentage: float
:param duration: SLA uptime duration override
:type duration: twitter.common.quantity.Amount
:rtype: set of unsafe hosts
"""
vector = self._client.sla_get_safe_domain_vector(
self.SLA_MIN_JOB_INSTANCE_COUNT, hostnames)
host_groups = vector.probe_hosts(percentage,
duration.as_(Time.SECONDS),
grouping_function)
unsafe_hostnames = set()
# Given that maintenance is performed 1 group at a time, any result longer than 1 group
# should be considered a batch failure.
if host_groups:
if len(host_groups) > 1:
log.error(
'Illegal multiple groups detected in SLA results. Skipping hosts: %s'
% hostnames)
return set(hostnames)
results, unsafe_hostnames = format_sla_results(host_groups,
unsafe_only=True)
if results:
print_results(results)
return unsafe_hostnames
return unsafe_hostnames
def sla_probe_hosts(cluster, percentage, duration):
"""usage: sla_probe_hosts
[--filename=FILENAME]
[--grouping=GROUPING]
[--hosts=HOSTS]
[--min_job_instance_count=COUNT]
cluster percentage duration
Probes individual hosts with respect to their job SLA.
Specifically, given a host, outputs all affected jobs with their projected SLAs
if the host goes down. In addition, if a job's projected SLA does not clear
the specified limits suggests the approximate time when that job reaches its SLA.
Output format:
HOST JOB PREDICTED_SLA SAFE? PREDICTED_SAFE_IN
where:
HOST - host being probed.
JOB - job that has tasks running on the host being probed.
PREDICTED_SLA - predicted effective percentage of up tasks if the host is shut down.
SAFE? - PREDICTED_SLA >= percentage
PREDICTED_SAFE_IN - expected wait time in seconds for the job to reach requested SLA threshold.
"""
options = app.get_options()
sla_percentage = parse_sla_percentage(percentage)
sla_duration = parse_time(duration)
hosts = parse_hostnames(options.filename, options.hosts)
get_grouping_or_die(options.grouping)
vector = AuroraClientAPI(CLUSTERS[cluster], options.verbosity).sla_get_safe_domain_vector(
options.min_instance_count, hosts
)
groups = vector.probe_hosts(sla_percentage, sla_duration.as_(Time.SECONDS), options.grouping)
output, _ = format_sla_results(groups)
print_results(output)
def sla_probe_hosts(cluster, percentage, duration):
"""usage: sla_probe_hosts
[--filename=FILENAME]
[--grouping=GROUPING]
[--hosts=HOSTS]
[--min_job_instance_count=COUNT]
cluster percentage duration
Probes individual hosts with respect to their job SLA.
Specifically, given a host, outputs all affected jobs with their projected SLAs
if the host goes down. In addition, if a job's projected SLA does not clear
the specified limits suggests the approximate time when that job reaches its SLA.
Output format:
HOST JOB PREDICTED_SLA SAFE? PREDICTED_SAFE_IN
where:
HOST - host being probed.
JOB - job that has tasks running on the host being probed.
PREDICTED_SLA - predicted effective percentage of up tasks if the host is shut down.
SAFE? - PREDICTED_SLA >= percentage
PREDICTED_SAFE_IN - expected wait time in seconds for the job to reach requested SLA threshold.
"""
options = app.get_options()
sla_percentage = parse_sla_percentage(percentage)
sla_duration = parse_time(duration)
hosts = parse_hostnames(options.filename, options.hosts)
get_grouping_or_die(options.grouping)
vector = AuroraClientAPI(
CLUSTERS[cluster],
options.verbosity).sla_get_safe_domain_vector(options.min_instance_count, hosts)
groups = vector.probe_hosts(sla_percentage, sla_duration.as_(Time.SECONDS), options.grouping)
output, _ = format_sla_results(groups)
print_results(output)
