def checkit(suffix, expectedMinLines):
logNameList = ["h2o_" + str(n.http_addr) + "_" + str(n.port) + suffix + ".log" for n in h2o_nodes.nodes]
lineCountList = []
for logName in logNameList:
lineCount = h2o_util.file_line_count(get_sandbox_name() + "/" + logName)
print logName, "lineCount:", lineCount
lineCountList.append(lineCount)
print logNameList
if len(h2o_nodes.nodes) != len(logNameList):
raise Exception("Should be %d logs, are %d" % len(h2o_nodes.nodes), len(logNameList))
# line counts seem to vary..check for "too small"
# variance in polling (cloud building and status)?
for i, l in enumerate(lineCountList):
if l < expectedMinLines:
raise Exception("node %d %s log is too small" % (i, logNameList[i]))
return (logNameList, lineCountList)
def log_download(self, logDir=None, timeoutSecs=30, **kwargs):
if logDir == None:
logDir = get_sandbox_name()
url = self.url('LogDownload.json')
log('Start ' + url);
print "\nDownloading h2o log(s) using:", url
r = requests.get(url, timeout=timeoutSecs, **kwargs)
if not r or not r.ok:
raise Exception("Maybe bad url? no r in log_download %s in %s:" % inspect.stack()[1][3])
z = zipfile.ZipFile(StringIO.StringIO(r.content))
print "z.namelist:", z.namelist()
print "z.printdir:", z.printdir()
nameList = z.namelist()
# the first is the h2ologs dir name.
h2oLogDir = logDir + "/" + nameList.pop(0)
print "h2oLogDir:", h2oLogDir
print "logDir:", logDir
# it's a zip of zipped files
# first unzip it
z = zipfile.ZipFile(StringIO.StringIO(r.content))
z.extractall(logDir)
# unzipped file should be in LOG_DIR now
# now unzip the files in that directory
for zname in nameList:
resultList = h2o_util.flat_unzip(logDir + "/" + zname, logDir)
print "\nlogDir:", logDir
for logfile in resultList:
numLines = sum(1 for line in open(logfile))
print logfile, "Lines:", numLines
print
return resultList
def build_cloud(node_count=1, base_port=None, hosts=None,
timeoutSecs=30, retryDelaySecs=1, cleanup=True, rand_shuffle=True,
conservative=False, create_json=False, clone_cloud=None,
init_sandbox=True, usecloud=False, usecloud_size=None, **kwargs):
# expectedSize is only used if usecloud
# usecloud can be passed thru build_cloud param, or command line
# not in config json though so no build_cloud_with_hosts path.
# redirect to build_cloud_with_json if a command line arg
# wants to force a test to ignore it's build_cloud/build_cloud_with_hosts
# (both come thru here)
# clone_cloud is just another way to get the effect (maybe ec2 config file thru
# build_cloud_with_hosts?
global stdout_wrapped
if not h2o_args.disable_time_stamp and not stdout_wrapped:
sys.stdout = OutWrapper(sys.stdout)
stdout_wrapped = True
if h2o_args.usecloud or usecloud:
# for now, just have fixed name in local file. (think of this as a temp or debug file)
# eventually we'll pass the json object instead for speed?
nodesJsonPathname = "h2o_fc-nodes.json"
elif h2o_args.clone_cloud_json:
nodesJsonPathname = h2o_args.clone_cloud_json
elif clone_cloud:
nodesJsonPathname = clone_cloud
else:
# normal build_cloud() doesn't use
nodesJsonPathname = None
# usecloud dominates over all
if (h2o_args.clone_cloud_json or clone_cloud) or (h2o_args.usecloud or usecloud):
# then build_cloud_with_json with json object
# we don't need to specify these defaults, but leave here to show that we can pass
# I suppose kwargs will have it
if h2o_args.usecloud:
ip_port = h2o_args.usecloud
elif usecloud:
ip_port = usecloud
else:
ip_port = None
# h2o_args dominates
if h2o_args.usecloud_size:
# only used for expected size
useCloudExpectedSize = h2o_args.usecloud_size
else:
useCloudExpectedSize = usecloud_size
nodesJsonObject = h2o_fc.find_cloud(ip_port=ip_port,
expectedSize=useCloudExpectedSize, nodesJsonPathname=nodesJsonPathname, **kwargs)
# potentially passed in kwargs
# hdfs_version='cdh4', hdfs_config=None, hdfs_name_node='172.16.1.176',
nodeList = build_cloud_with_json(h2o_nodes_json=nodesJsonPathname)
return nodeList
# else
# moved to here from unit_main. so will run with nosetests too!
# Normally do this.
# Don't if build_cloud_with_hosts() did and put a flatfile in there already!
if init_sandbox:
clean_sandbox()
log("#*********************************************************************")
log("Starting new test: " + h2o_args.python_test_name + " at build_cloud() ")
log("#*********************************************************************")
# start up h2o to report the java version (once). output to python stdout
# only do this for regression testing
# temporarily disable this, to go a little faster
# if getpass.getuser() == 'jenkins':
# check_h2o_version()
ports_per_node = 2
nodeList = []
# shift the port used to run groups of tests on the same machine at the same time?
base_port = get_base_port(base_port)
try:
# if no hosts list, use psutil method on local host.
totalNodes = 0
# doing this list outside the loops so we can shuffle for better test variation
# this jvm startup shuffle is independent from the flatfile shuffle
portList = [base_port + ports_per_node * i for i in range(node_count)]
if hosts is None:
# if use_flatfile, we should create it
# because tests will just call build_cloud with use_flatfile=True
# best to just create it all the time..may or may not be used
write_flatfile(node_count=node_count, base_port=base_port)
hostCount = 1
if rand_shuffle:
random.shuffle(portList)
for p in portList:
verboseprint("psutil starting node", i)
newNode = LocalH2O(port=p, node_id=totalNodes, **kwargs)
nodeList.append(newNode)
totalNodes += 1
else:
# if hosts, the flatfile was created and uploaded to hosts already
# I guess don't recreate it, don't overwrite the one that was copied beforehand.
# we don't always use the flatfile (use_flatfile=False)
# Suppose we could dispatch from the flatfile to match it's contents
# but sometimes we want to test with a bad/different flatfile then we invoke h2o?
hostCount = len(hosts)
hostPortList = []
for h in hosts:
for port in portList:
hostPortList.append((h, port))
if rand_shuffle: random.shuffle(hostPortList)
for (h, p) in hostPortList:
verboseprint('ssh starting node', totalNodes, 'via', h)
newNode = h.remote_h2o(port=p, node_id=totalNodes, **kwargs)
nodeList.append(newNode)
totalNodes += 1
verboseprint("Attempting Cloud stabilize of", totalNodes, "nodes on", hostCount, "hosts")
start = time.time()
# UPDATE: best to stabilize on the last node!
stabilize_cloud(nodeList[0], nodeList,
timeoutSecs=timeoutSecs, retryDelaySecs=retryDelaySecs, noSandboxErrorCheck=True)
verboseprint(len(nodeList), "Last added node stabilized in ", time.time() - start, " secs")
verboseprint("Built cloud: %d nodes on %d hosts, in %d s" % \
(len(nodeList), hostCount, (time.time() - start)))
h2p.red_print("Built cloud:", nodeList[0].java_heap_GB, "GB java heap(s) with",
len(nodeList), "total nodes")
# FIX! using "consensus" in node[-1] should mean this is unnecessary?
# maybe there's a bug. For now do this. long term: don't want?
# UPDATE: do it for all cases now 2/14/13
if conservative: # still needed?
for n in nodeList:
stabilize_cloud(n, nodeList, timeoutSecs=timeoutSecs, noSandboxErrorCheck=True)
# this does some extra checking now
# verifies cloud name too if param is not None
verify_cloud_size(nodeList, expectedCloudName=nodeList[0].cloud_name)
# best to check for any errors due to cloud building right away?
check_sandbox_for_errors(python_test_name=h2o_args.python_test_name)
except:
# nodeList might be empty in some exception cases?
# no shutdown issued first, though
if cleanup and nodeList:
for n in nodeList: n.terminate()
check_sandbox_for_errors(python_test_name=h2o_args.python_test_name)
raise
print len(nodeList), "total jvms in H2O cloud"
# put the test start message in the h2o log, to create a marker
nodeList[0].h2o_log_msg()
if h2o_args.config_json:
LOG_DIR = get_sandbox_name()
# like cp -p. Save the config file, to sandbox
print "Saving the ", h2o_args.config_json, "we used to", LOG_DIR
shutil.copy(h2o_args.config_json, LOG_DIR + "/" + os.path.basename(h2o_args.config_json))
# Figure out some stuff about how this test was run
cs_time = str(datetime.datetime.now())
cs_cwd = os.getcwd()
cs_python_cmd_line = "python %s %s" % (h2o_args.python_test_name, h2o_args.python_cmd_args)
cs_python_test_name = h2o_args.python_test_name
if h2o_args.config_json:
cs_config_json = os.path.abspath(h2o_args.config_json)
else:
cs_config_json = None
cs_username = h2o_args.python_username
cs_ip = h2o_args.python_cmd_ip
# dump the nodes state to a json file # include enough extra info to have someone
# rebuild the cloud if a test fails that was using that cloud.
if create_json:
q = {
'cloud_start':
{
'time': cs_time,
'cwd': cs_cwd,
'python_test_name': cs_python_test_name,
'python_cmd_line': cs_python_cmd_line,
'config_json': cs_config_json,
'username': cs_username,
'ip': cs_ip,
},
'h2o_nodes': h2o_util.json_repr(nodeList),
}
with open('h2o-nodes.json', 'w+') as f:
f.write(json.dumps(q, indent=4))
# save it to a local global copy, in case it's needed for tearDown
h2o_nodes.nodes[:] = nodeList
return nodeList
clean_sandbox_doneToLine,
verboseprint,
OutWrapper,
log,
flatfile_pathname,
dump_json,
find_file,
check_h2o_version,
)
from h2o_objects import LocalH2O, RemoteH2O, ExternalH2O
import h2o_fc
import h2o_hosts
# print "h2o_bc"
LOG_DIR = get_sandbox_name()
# ************************************************************
def default_hosts_file():
if os.environ.has_key("H2O_HOSTS_FILE"):
return os.environ["H2O_HOSTS_FILE"]
return "pytest_config-{0}.json".format(getpass.getuser())
# ************************************************************
# node_count is number of H2O instances per host if hosts is specified.
# hack: this returns true for the --usecloud/-uc cases, to force it thru
# build_cloud/build_cloud_with_json/find_cloud. also for the -ccj cases
def decide_if_localhost():
if h2o_args.usecloud:
print "* Will ask h2o node about cloud using -uc argument:", h2o_args.usecloud
import h2o_args
import h2o_nodes
import h2o_print as h2p, h2o_util
import h2o_import as h2i
from h2o_test import \
get_sandbox_name, clean_sandbox, check_sandbox_for_errors, clean_sandbox_doneToLine,\
verboseprint, OutWrapper, log, flatfile_pathname, dump_json, find_file, check_h2o_version
from h2o_objects import LocalH2O, RemoteH2O, ExternalH2O
import h2o_fc
import h2o_hosts
# print "h2o_bc"
LOG_DIR = get_sandbox_name()
#************************************************************
def default_hosts_file():
if os.environ.has_key("H2O_HOSTS_FILE"):
return os.environ["H2O_HOSTS_FILE"]
return 'pytest_config-{0}.json'.format(getpass.getuser())
#************************************************************
# node_count is number of H2O instances per host if hosts is specified.
# hack: this returns true for the --usecloud/-uc cases, to force it thru
# build_cloud/build_cloud_with_json/find_cloud. also for the -ccj cases
def decide_if_localhost():
if h2o_args.usecloud:
print "* Will ask h2o node about cloud using -uc argument:", h2o_args.usecloud
return True