def osdids():
ret, outbuf, outs = json_command(cluster_handle, prefix='osd ls')
if ret == -errno.EINVAL:
# try old mon
ret, outbuf, outs = send_command(cluster_handle, cmd=['osd', 'ls'])
if ret:
raise RuntimeError('Can\'t contact mon for osd list')
return [i for i in outbuf.split('\n') if i != '']
def ceph_rbd_du():
global rbd_du_out,rbd_du_out_ofs
while 1:
starttime = datetime.datetime.now()
outbuf = ""
target = ('mon', '')
timeout=0
verbose=False
inbuf=''
cmddict = {}
cmddict.update({'prefix': u'osd lspools', 'format': 'json'})
ret, outbuf, outs = send_command(cluster, target, [json.dumps(cmddict)],inbuf, timeout, verbose)
lspools = json.loads(outbuf)
poollist = []
for item in lspools:
d = {}
#print item["poolname"]
cmdstr = "rbd -p "+ item["poolname"] + " du"
#print cmdstr
images = []
status, outbuf = commands.getstatusoutput(cmdstr)
outarray = outbuf.split('\n')
del outarray[0]
if ( len(outarray) < 1 ):
continue
del outarray[len(outarray)-1]
for index,value in enumerate(outarray):
a = value.split(' ')
tmparray = []
for index,tmpstr in enumerate(a):
if ( len(tmpstr) >= 1 ):
tmparray.append(tmpstr)
image = {}
image["used"] = tmparray[2]
image["provisioned"] = tmparray[1]
image["name"] = tmparray[0]
images.append(image)
d["images"] = images
d["poolname"] = item["poolname"]
poollist.append(d)
mutex.acquire()
rbd_du_out_ofs=0
rbd_du_out = "begin"+json.dumps(poollist)+"end"
mutex.release()
#print rbd_du_out
while( rbd_du_out_ofs < len(rbd_du_out) ):
time.sleep(1)
endtime = datetime.datetime.now()
#print (endtime - starttime).seconds
if ( (endtime - starttime).seconds < 3600 ):
time.sleep( 3600 - (endtime - starttime).seconds )
def monids():
ret, outbuf, outs = json_command(cluster_handle, prefix='mon dump',
argdict={'format':'json'})
if ret == -errno.EINVAL:
# try old mon
ret, outbuf, outs = send_command(cluster_handle,
cmd=['mon', 'dump', '--format=json'])
if ret:
raise RuntimeError('Can\'t contact mon for mon list')
d = json.loads(outbuf)
return [m['name'] for m in d['mons']]
def mdsids():
ret, outbuf, outs = json_command(cluster_handle, prefix='mds dump',
argdict={'format':'json'})
if ret == -errno.EINVAL:
# try old mon
ret, outbuf, outs = send_command(cluster_handle,
cmd=['mds', 'dump', '--format=json'])
if ret:
raise RuntimeError('Can\'t contact mon for mds list')
d = json.loads(outbuf)
l = []
infodict = d['info']
for mdsdict in infodict.values():
l.append(mdsdict['name'])
return l
def send_cmd(cluster, target=('mon', ''), cmd=None, inbuf=b'', timeout=0, verbose=False):
try:
ret, outbuf, outs = send_command(cluster, target, cmd, inbuf, timeout, verbose)
except Exception as e:
print e
return ret, outbuf, outs
def main():
# 这里是从环境变量中拿CEPH_ARGS
# 因为这里存在一种用法：比如把日志打印到stderr
# export CEPH_ARGS="--log-to-stderr"
ceph_args = os.environ.get('CEPH_ARGS')
# 如果这里有针对ceph的参数，那么需要进行处理。
# 一般而言，这里是没有这些参数的。
if ceph_args:
if "injectargs" in sys.argv:
i = sys.argv.index("injectargs")
sys.argv = sys.argv[:i] + ceph_args.split() + sys.argv[i:]
else:
sys.argv.extend(ceph_args.split())
parser, parsed_args, childargs = parse_cmdargs()
# 如果只是查看版本，那么直接输出版本，然后退出。
if parsed_args.version:
print 'ceph version {0} ({1})'.format(CEPH_GIT_NICE_VER, CEPH_GIT_VER)
return 0
# verbose是进入一种交互式的命令行方式，类似于直接运行python。
global verbose
verbose = parsed_args.verbose
if verbose:
print >> sys.stderr, "parsed_args: {0}, childargs: {1}".format(parsed_args, childargs)
# 这里只是一个客户端工具，是不能加上--admin-socket参数来运行。
# 守护进程，比如ceph-mon，ceph-osd才可以用这种方式来运行。
if parsed_args.admin_socket_nope:
print >> sys.stderr, '--admin-socket is used by daemons; '\
'you probably mean --admin-daemon/daemon'
return 1
# pass on --id, --name, --conf
name = 'client.admin'
# 这里获取需要认证的客户端用户名。一般是采用client.admin。
# 如果用了id，那么就直接用id。
if parsed_args.client_id:
name = 'client.' + parsed_args.client_id
# 如果给了名字，那么就直接使用名字。
if parsed_args.client_name:
name = parsed_args.client_name
# default '' means default conf search
# 这里设置使用的配置文件，一般是/etc/ceph/ceph.conf。
conffile = ''
if parsed_args.cephconf:
conffile = parsed_args.cephconf
# For now, --admin-daemon is handled as usual. Try it
# first in case we can't connect() to the cluster
# 这里设置输出格式。
format = parsed_args.output_format
# 这里设置socket。在运行ceph -s的时候这段代码没什么用，不会用到。
sockpath = None
if parsed_args.admin_socket:
sockpath = parsed_args.admin_socket
elif len(childargs) > 0 and childargs[0] == "daemon":
# Treat "daemon <path>" or "daemon <name>" like --admin_daemon <path>
if len(childargs) > 2:
if childargs[1].find('/') >= 0:
sockpath = childargs[1]
else:
# try resolve daemon name
try:
sockpath = ceph_conf(parsed_args, 'admin_socket',
childargs[1])
except Exception as e:
print >> sys.stderr, \
'Can\'t get admin socket path: ' + str(e)
return errno.EINVAL
# for both:
childargs = childargs[2:]
else:
print >> sys.stderr, 'daemon requires at least 3 arguments'
return errno.EINVAL
# ceph -s的时候，会直到这里来执行。
if sockpath:
try:
print admin_socket(sockpath, childargs, format)
except Exception as e:
print >> sys.stderr, 'admin_socket: {0}'.format(e)
return errno.EINVAL
return 0
timeout = None
if parsed_args.cluster_timeout:
timeout = parsed_args.cluster_timeout
# basic help
if parsed_args.help:
do_basic_help(parser, childargs)
# handle any 'generic' ceph arguments that we didn't parse here
global cluster_handle
# rados.Rados() will call rados_create2, and then read the conf file,
# and then set the keys from the dict. So we must do these
# "pre-file defaults" first (see common_preinit in librados)
conf_defaults = {
'log_to_stderr':'true',
'err_to_stderr':'true',
'log_flush_on_exit':'true',
}
if 'injectargs' in childargs:
position = childargs.index('injectargs')
injectargs = childargs[position:]
childargs = childargs[:position]
if verbose:
print >> sys.stderr, 'Separate childargs {0} from injectargs {1}'.\
format(childargs, injectargs)
else:
injectargs = None
clustername = 'ceph'
if parsed_args.cluster:
clustername = parsed_args.cluster
try:
cluster_handle = rados.Rados(name=name, clustername=clustername,
conf_defaults=conf_defaults,
conffile=conffile)
retargs = cluster_handle.conf_parse_argv(childargs)
except rados.Error as e:
print >> sys.stderr, 'Error initializing cluster client: {0}'.\
format(repr(e))
return 1
childargs = retargs
if not childargs:
childargs = []
# -- means "stop parsing args", but we don't want to see it either
if '--' in childargs:
childargs.remove('--')
if injectargs and '--' in injectargs:
injectargs.remove('--')
# special deprecation warning for 'ceph <type> tell'
# someday 'mds' will be here too
if len(childargs) >= 2 and \
childargs[0] in ['mon', 'osd'] and \
childargs[1] == 'tell':
print >> sys.stderr, '"{0} tell" is deprecated; try "tell {0}.<id>" instead (id can be "*") '.format(childargs[0])
return 1
if parsed_args.help:
# short default timeout for -h
if not timeout:
timeout = 5
hdr('Monitor commands:')
print '[Contacting monitor, timeout after %d seconds]' % timeout
if childargs and childargs[0] == 'ping':
if len(childargs) < 2:
print >> sys.stderr, '"ping" requires a monitor name as argument: "ping mon.<id>"'
return 1
try:
if childargs and childargs[0] == 'ping':
return ping_monitor(cluster_handle, childargs[1])
cluster_handle.connect(timeout=timeout)
except KeyboardInterrupt:
print >> sys.stderr, 'Cluster connection aborted'
return 1
except Exception as e:
print >> sys.stderr, 'Error connecting to cluster: {0}'.\
format(e.__class__.__name__)
return 1
if parsed_args.help:
return do_extended_help(parser, childargs)
# implement -w/--watch_*
# This is ugly, but Namespace() isn't quite rich enough.
level = ''
for k, v in parsed_args._get_kwargs():
if k.startswith('watch') and v:
if k == 'watch':
level = 'info'
else:
level = k.replace('watch_', '')
if level:
# an awfully simple callback
def watch_cb(arg, line, who, stamp_sec, stamp_nsec, seq, level, msg):
print line
sys.stdout.flush()
# first do a ceph status
ret, outbuf, outs = json_command(cluster_handle, prefix='status')
if ret == -errno.EINVAL:
# try old mon
ret, outbuf, outs = send_command(cluster_handle, cmd=['status'])
# old mon returns status to outs...ick
if ret == 0:
outbuf += outs
if ret:
print >> sys.stderr, "status query failed: ", outs
return ret
print outbuf
# this instance keeps the watch connection alive, but is
# otherwise unused
logwatch = rados.MonitorLog(cluster_handle, level, watch_cb, 0)
# loop forever letting watch_cb print lines
try:
signal.pause()
except KeyboardInterrupt:
# or until ^C, at least
return 0
# read input file, if any
inbuf = ''
if parsed_args.input_file:
try:
with open(parsed_args.input_file, 'r') as f:
inbuf = f.read()
except Exception as e:
print >> sys.stderr, 'Can\'t open input file {0}: {1}'.format(parsed_args.input_file, e)
return 1
# prepare output file, if any
if parsed_args.output_file:
try:
outf = open(parsed_args.output_file, 'w')
except Exception as e:
print >> sys.stderr, \
'Can\'t open output file {0}: {1}'.\
format(parsed_args.output_file, e)
return 1
# -s behaves like a command (ceph status).
if parsed_args.status:
childargs.insert(0, 'status')
try:
target = find_cmd_target(childargs)
except Exception as e:
print >> sys.stderr, \
'error handling command target: {0}'.format(e)
return 1
# Repulsive hack to handle tell: lop off 'tell' and target
# and validate the rest of the command. 'target' is already
# determined in our callers, so it's ok to remove it here.
is_tell = False
if len(childargs) and childargs[0] == 'tell':
childargs = childargs[2:]
is_tell = True
if is_tell:
if injectargs:
childargs = injectargs
if not len(childargs):
print >> sys.stderr, \
'Cannot use \'tell\' with interactive mode'
return errno.EINVAL
# fetch JSON sigs from command
# each line contains one command signature (a placeholder name
# of the form 'cmdNNN' followed by an array of argument descriptors)
# as part of the validated argument JSON object
targets = [target]
if target[1] == '*':
if target[0] == 'osd':
targets = [(target[0], o) for o in osdids()]
elif target[0] == 'mon':
targets = [(target[0], m) for m in monids()]
final_ret = 0
for target in targets:
# prettify? prefix output with target, if there was a wildcard used
prefix = ''
suffix = ''
if not parsed_args.output_file and len(targets) > 1:
prefix = '{0}.{1}: '.format(*target)
suffix = '\n'
ret, outbuf, outs = json_command(cluster_handle, target=target,
prefix='get_command_descriptions')
compat = False
if ret == -errno.EINVAL:
# send command to old monitor or OSD
if verbose:
print prefix + '{0} to old {1}'.format(' '.join(childargs), target[0])
compat = True
if parsed_args.output_format:
childargs.extend(['--format', parsed_args.output_format])
ret, outbuf, outs = send_command(cluster_handle, target, childargs,
inbuf)
if ret == -errno.EINVAL:
# did we race with a mon upgrade? try again!
ret, outbuf, outs = json_command(cluster_handle, target=target,
prefix='get_command_descriptions')
if ret == 0:
compat = False # yep, carry on
if not compat:
if ret:
if ret < 0:
outs = 'problem getting command descriptions from {0}.{1}'.format(*target)
else:
sigdict = parse_json_funcsigs(outbuf, 'cli')
if parsed_args.completion:
return complete(sigdict, childargs, target)
ret, outbuf, outs = new_style_command(parsed_args, childargs, target,
sigdict, inbuf, verbose)
# debug tool: send any successful command *again* to
# verify that it is idempotent.
if not ret and 'CEPH_CLI_TEST_DUP_COMMAND' in os.environ:
ret, outbuf, outs = new_style_command(parsed_args, childargs, target,
sigdict, inbuf, verbose)
if ret < 0:
ret = -ret
print >> sys.stderr, prefix + 'Second attempt of previously successful command failed with {0}: {1}'.format(errno.errorcode[ret], outs)
if ret < 0:
ret = -ret
print >> sys.stderr, prefix + 'Error {0}: {1}'.format(errno.errorcode[ret], outs)
if len(targets) > 1:
final_ret = ret
else:
return ret
# this assumes outs never has useful command output, only status
if compat:
if ret == 0:
# old cli/mon would send status string to stdout on non-error
print outs
else:
if outs:
print >> sys.stderr, prefix + outs
if (parsed_args.output_file):
outf.write(outbuf)
else:
# hack: old code printed status line before many json outputs
# (osd dump, etc.) that consumers know to ignore. Add blank line
# to satisfy consumers that skip the first line, but not annoy
# consumers that don't.
if parsed_args.output_format and \
parsed_args.output_format.startswith('json') and \
not compat:
sys.stdout.write('\n')
# if we are prettifying things, normalize newlines. sigh.
if suffix != '':
outbuf = outbuf.rstrip()
if outbuf != '':
sys.stdout.write(prefix + outbuf + suffix)
sys.stdout.flush()
if (parsed_args.output_file):
outf.close()
if final_ret:
return final_ret
return 0
def main(argv):
global warn
mutex = threading.Lock()
try:
cluster = rados.Rados(conffile='')
except TypeError as e:
print 'Argument validation error: ', e
raise e
print "Created cluster handle."
try:
cluster.connect()
except Exception as e:
print "connection error: ", e
raise e
finally:
print "Connected to the cluster."
s = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
try:
os.remove( socket_path )
except :
pass
#s.bind( socket_path )
try:
s.bind( socket_path )
except :
#os._exit(0)
pass
#uid = pwd.getpwnam("zabbix")
uid = pwd.getpwnam(argv[1])
os.chown(socket_path, uid.pw_uid, uid.pw_gid)
s.listen(1)
def ceph_rbd_du():
global rbd_du_out
while 1:
starttime = datetime.datetime.now()
outbuf = ""
target = ('mon', '')
timeout=0
verbose=False
inbuf=''
cmddict = {}
cmddict.update({'prefix': u'osd lspools', 'format': 'json'})
ret, outbuf, outs = send_command(cluster, target, [json.dumps(cmddict)],inbuf, timeout, verbose)
lspools = json.loads(outbuf)
poollist = []
for item in lspools:
d = {}
#print item["poolname"]
cmdstr = "rbd -p "+ item["poolname"] + " du"
#print cmdstr
images = []
status, outbuf = commands.getstatusoutput(cmdstr)
outarray = outbuf.split('\n')
del outarray[0]
del outarray[len(outarray)-1]
for index,value in enumerate(outarray):
a = value.split(' ')
tmparray = []
for index,tmpstr in enumerate(a):
if ( len(tmpstr) >= 1 ):
tmparray.append(tmpstr)
image = {}
image["used"] = tmparray[2]
image["provisioned"] = tmparray[1]
image["name"] = tmparray[0]
images.append(image)
d["images"] = images
d["poolname"] = item["poolname"]
poollist.append(d)
mutex.acquire()
rbd_du_out = json.dumps(poollist)
mutex.release()
#print rbd_du_out
endtime = datetime.datetime.now()
#print (endtime - starttime).seconds
if ( (endtime - starttime).seconds < 300 ):
time.sleep( 300 - (endtime - starttime).seconds )
t = threading.Thread(target=ceph_rbd_du)
t.setDaemon(True)
t.start()
def watch_cb(arg, line, who, stamp_sec, stamp_nsec, seq, level, msg):
#print(line)
#sys.stdout.flush()
find_warn(line)
mutex.acquire()
cluster_log_list.insert(0,line)
mutex.release()
# this instance keeps the watch connection alive, but is
# otherwise unused
level = "info"
run_in_thread(cluster.monitor_log, level, watch_cb, 0)
warn = "0"
while 1:
conn, addr = s.accept()
cmd = conn.recv(1024)
if not cmd: break
print cmd
outbuf = ""
target = ('mon', '')
timeout=0
verbose=False
inbuf=''
cmddict = {}
if ( cmd == "rbd_du" ):
print rbd_du_out
mutex.acquire()
conn.send(rbd_du_out)
mutex.release()
conn.close()
continue
elif ( cmd == "log" ):
outbuf = ''
mutex.acquire()
for log_str in cluster_log_list:
outbuf += log_str + "\n"
del cluster_log_list[:]
mutex.release()
conn.send(outbuf)
conn.close()
continue
elif ( cmd == "warn" ):
conn.send( str(warn) )
conn.close()
continue
elif ( cmd == "status" ):
cmddict.update({'prefix': u'status', 'format': 'json'})
elif( cmd == "osd_pool_stats" ):
cmddict.update({'prefix': u'osd pool stats', 'format': 'json'})
elif ( cmd == "osd_tree" ):
cmddict.update({'prefix': u'osd tree', 'format': 'json'})
elif ( cmd == "osd_df" ):
cmddict.update({'prefix': u'osd df', 'format': 'json'})
elif (cmd == "pg_stat" ):
cmddict.update({'prefix': u'pg stat', 'format': 'json'})
elif (cmd == "osd_perf" ):
cmddict.update({'prefix': u'osd perf', 'format': 'json'})
elif (cmd == "df" ):
cmddict.update({'prefix': u'df', 'format': 'json'})
ret, outbuf, outs = send_command(cluster, target, [json.dumps(cmddict)],inbuf, timeout, verbose)
find_warn(outbuf)
conn.send(outbuf)
conn.close()
