[multi-asic] Enhancing monit process checker for multi-asic. (#6100)

Added Support of process checker for work on multi-asic platforms.
This commit is contained in:
abdosi 2020-12-04 10:39:43 -08:00 committed by GitHub
parent 8a04487499
commit 59c1e3a78a
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 50 additions and 11 deletions

View File

@ -1,10 +1,12 @@
#!/usr/bin/python3
import argparse
import ast
import sys
import syslog
import psutil
from sonic_py_common import multi_asic
import swsssdk
@ -27,20 +29,55 @@ def check_process_existence(container_name, process_cmdline):
# We leveraged the psutil library to help us check whether the process is running or not.
# If the process entity is found in process tree and it is also in the 'running' or 'sleeping'
# state, then it will be marked as 'running'.
is_running = False
for process in psutil.process_iter(["cmdline", "status"]):
# For given feature we get the host and network namespace instances it's processes should be running
# based on it's scope and add it to expected set.
# From psutil we get number of running instances of the processes and add it to the the actual set
# Difference bwetween expected and actual set provides instances where the processes are not running
# and will be logged as syslog message by monit
process_namespace_expected_set = set()
process_namespace_found_set = set()
has_global_scope = ast.literal_eval(feature_table[container_name].get('has_global_scope', 'True'))
has_per_asic_scope = ast.literal_eval(feature_table[container_name].get('has_per_asic_scope', 'False'))
if has_global_scope:
process_namespace_expected_set.add(multi_asic.DEFAULT_NAMESPACE)
if has_per_asic_scope:
process_namespace_expected_set.update(multi_asic.get_namespace_list())
for process in psutil.process_iter(["cmdline", "status", "pid"]):
try:
if ((' '.join(process.cmdline())).startswith(process_cmdline) and process.status() in ["running", "sleeping"]):
is_running = True
break
process_namespace_found_set.add(multi_asic.get_current_namespace(process.info['pid']))
except psutil.NoSuchProcess:
pass
if not is_running:
# If this script is run by Monit, then the following output will be appended to
# Monit's syslog message.
print("'{}' is not running.".format(process_cmdline))
sys.exit(1)
process_namespace_diff_set = process_namespace_expected_set.difference(process_namespace_found_set)
if process_namespace_diff_set:
host_display_str = ""
namespace_display_str = ""
for ns in process_namespace_diff_set:
if ns == multi_asic.DEFAULT_NAMESPACE:
host_display_str = " in host"
else:
if not namespace_display_str:
namespace_display_str = " in namespace " + ns
else:
namespace_display_str += ", " + ns
join_str = " and" if host_display_str and namespace_display_str else ""
# If this script is run by Monit, then the following output will be appended to
# Monit's syslog message.
print("'{}' is not running{}{}{}".format(process_cmdline, host_display_str, join_str, namespace_display_str))
sys.exit(1)
else:
syslog.syslog(syslog.LOG_ERR, "container '{}' is not included in SONiC image or the given container name is invalid!"
.format(container_name))

View File

@ -138,14 +138,14 @@ def get_asic_id_from_name(asic_name):
raise ValueError('Unknown asic namespace name {}'.format(asic_name))
def get_current_namespace():
def get_current_namespace(pid=None):
"""
This API returns the network namespace in which it is
invoked. In case of global namepace the API returns None
"""
net_namespace = None
command = ["/bin/ip netns identify", str(os.getpid())]
command = ["sudo /bin/ip netns identify {}".format(os.getpid() if not pid else pid)]
proc = subprocess.Popen(command,
stdout=subprocess.PIPE,
shell=True,
@ -159,6 +159,8 @@ def get_current_namespace():
)
if stdout.rstrip('\n') != "":
net_namespace = stdout.rstrip('\n')
else:
net_namespace = DEFAULT_NAMESPACE
except OSError as e:
raise OSError("Error running command {}".format(command))