def wait_for_server_start(pidFile, scmStatus):
properties = get_ambari_properties()
if properties == -1:
err = "Error getting ambari properties"
raise FatalException(-1, err)
#wait for server process for SERVER_START_TIMEOUT seconds
sys.stdout.write('Waiting for server start...')
sys.stdout.flush()
pids = []
server_started = False
# looking_for_pid() might return partrial pid list on slow hardware
for i in range(1, SERVER_START_RETRIES):
pids = looking_for_pid(SERVER_SEARCH_PATTERN, SERVER_START_TIMEOUT)
sys.stdout.write('\n')
sys.stdout.flush()
if save_main_pid_ex(
pids, pidFile,
locate_all_file_paths('sh', '/bin') +
locate_all_file_paths('bash', '/bin') +
locate_all_file_paths('dash', '/bin'), IS_FOREGROUND):
server_started = True
break
else:
sys.stdout.write("Unable to determine server PID. Retrying...\n")
sys.stdout.flush()
exception = None
if server_started:
ambari_server_ui_port = get_ambari_server_ui_port(properties)
if not wait_for_ui_start(int(ambari_server_ui_port), WEB_UI_INIT_TIME):
exception = FatalException(1, "Server not yet listening on http port " + ambari_server_ui_port + \
" after " + str(WEB_UI_INIT_TIME) + " seconds. Exiting.")
elif get_live_pids_count(pids) <= 0:
exitcode = check_exitcode(
os.path.join(configDefaults.PID_DIR, EXITCODE_NAME))
exception = FatalException(
-1,
AMBARI_SERVER_DIE_MSG.format(exitcode,
configDefaults.SERVER_OUT_FILE))
else:
exception = FatalException(-1, AMBARI_SERVER_NOT_STARTED_MSG)
if 'Database consistency check: failed' in open(
configDefaults.SERVER_OUT_FILE).read():
print "DB configs consistency check failed. Run \"ambari-server start --skip-database-check\" to skip. " \
"You may try --auto-fix-database flag to attempt to fix issues automatically. " \
"If you use this \"--skip-database-check\" option, do not make any changes to your cluster topology " \
"or perform a cluster upgrade until you correct the database consistency issues. See " + \
configDefaults.DB_CHECK_LOG + " for more details on the consistency issues."
elif 'Database consistency check: warning' in open(
configDefaults.SERVER_OUT_FILE).read():
print "DB configs consistency check found warnings. See " + configDefaults.DB_CHECK_LOG + " for more details."
else:
print "DB configs consistency check: no errors and warnings were found."
if exception:
raise exception
def wait_for_pid(pids, server_init_timeout, occupy_port_timeout,
init_web_ui_timeout, server_out_file, db_check_log,
properties):
"""
Check pid for existence during timeout
"""
from ambari_server.serverConfiguration import get_ambari_server_ui_port
ambari_server_ui_port = int(get_ambari_server_ui_port(properties))
server_ui_port_occupied = False
tstart = time.time()
pid_live = 0
while int(time.time() - tstart) <= occupy_port_timeout and len(pids) > 0:
sys.stdout.write('.')
sys.stdout.flush()
pid_live = 0
for item in pids:
if pid_exists(item["pid"]):
pid_live += 1
time.sleep(1)
try:
sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
sock.settimeout(1)
sock.connect(('localhost', ambari_server_ui_port))
print "\nServer started listening on " + str(ambari_server_ui_port)
server_ui_port_occupied = True
break
except Exception as e:
#print str(e)
pass
if 'Database consistency check: failed' in open(server_out_file).read():
print "\nDB configs consistency check failed. Run \"ambari-server start --skip-database-check\" to skip. " \
"If you use this \"--skip-database-check\" option, do not make any changes to your cluster topology " \
"or perform a cluster upgrade until you correct the database consistency issues. See " + \
db_check_log + "for more details on the consistency issues."
elif 'Database consistency check: warning' in open(server_out_file).read():
print "\nDB configs consistency check found warnings. See " + db_check_log + " for more details."
else:
print "\nDB configs consistency check: no errors and warnings were found."
if not server_ui_port_occupied:
raise FatalException(
1, "Server not yet listening on http port " +
str(ambari_server_ui_port) + " after " +
str(occupy_port_timeout + server_init_timeout) +
" seconds. Exiting.")
tstart = time.time()
print "Waiting for 10 seconds, for server WEB UI initialization"
while int(time.time() - tstart) <= init_web_ui_timeout and len(pids) > 0:
sys.stdout.write('.')
sys.stdout.flush()
pid_live = 0
for item in pids:
if pid_exists(item["pid"]):
pid_live += 1
time.sleep(1)
return pid_live
def wait_for_server_start(pidFile, scmStatus):
properties = get_ambari_properties()
if properties == -1:
err = "Error getting ambari properties"
raise FatalException(-1, err)
#wait for server process for SERVER_START_TIMEOUT seconds
sys.stdout.write('Waiting for server start...')
sys.stdout.flush()
pids = []
pid = None
# looking_for_pid() might return partrial pid list on slow hardware
for i in range(1, SERVER_START_RETRIES):
pids = looking_for_pid(SERVER_SEARCH_PATTERN, SERVER_START_TIMEOUT)
pid = save_main_pid_ex(
pids, pidFile,
locate_all_file_paths('sh', '/bin') +
locate_all_file_paths('bash', '/bin') +
locate_all_file_paths('dash', '/bin'), IS_FOREGROUND)
if pid:
break
else:
sys.stdout.write("Unable to determine server PID. Retrying...\n")
sys.stdout.flush()
exception = None
if pid:
ambari_server_ui_port = get_ambari_server_ui_port(properties)
web_server_startup_timeout = get_web_server_startup_timeout(properties)
waitStart = time.time()
if not wait_for_ui_start(int(ambari_server_ui_port), pid,
web_server_startup_timeout):
waitTime = int(time.time() - waitStart)
# Java process stopped, due to a DB check or other startup issue
if waitTime < web_server_startup_timeout:
exception = FatalException(-1, AMBARI_SERVER_STOPPED)
# UI didn't come up on time
else:
exception = FatalException(
1,
AMBARI_SERVER_UI_TIMEOUT.format(
ambari_server_ui_port, web_server_startup_timeout))
elif get_live_pids_count(pids) <= 0:
exitcode = check_exitcode(
os.path.join(configDefaults.PID_DIR, EXITCODE_NAME))
exception = FatalException(
-1,
AMBARI_SERVER_DIE_MSG.format(exitcode,
configDefaults.SERVER_OUT_FILE))
else:
exception = FatalException(-1, AMBARI_SERVER_NOT_STARTED_MSG)
if os.path.isfile(configDefaults.SERVER_OUT_FILE):
if 'DB_CHECK_ERROR' in open(configDefaults.SERVER_OUT_FILE).read():
print "\nDB configs consistency check failed. Run \"ambari-server start --skip-database-check\" to skip. " \
"You may try --auto-fix-database flag to attempt to fix issues automatically. " \
"If you use this \"--skip-database-check\" option, do not make any changes to your cluster topology " \
"or perform a cluster upgrade until you correct the database consistency issues. See " + \
configDefaults.DB_CHECK_LOG + " for more details on the consistency issues."
elif 'DB_CHECK_WARNING' in open(configDefaults.SERVER_OUT_FILE).read():
print "\nDB configs consistency check found warnings. See " + configDefaults.DB_CHECK_LOG + " for more details."
# Only presume that DB check was successful if it explicitly appears in the log. An unexpected error may prevent
# the consistency check from running at all, so missing error/warning message in the log cannot imply the check was
# successful
elif 'DB_CHECK_SUCCESS' in open(configDefaults.SERVER_OUT_FILE).read():
print "\nDB configs consistency check: no errors and warnings were found."
else:
sys.stdout.write(configDefaults.SERVER_OUT_FILE + " does not exist")
if exception:
raise exception