Révision a80cf00c
| b/papers/2014/reservation/experiments/src/python/lib/kadeploy.py | ||
|---|---|---|
| 19 | 19 |
|
| 20 | 20 |
class DeploymentStat: |
| 21 | 21 |
|
| 22 |
def __init__(self, host, start, step1_dur, step2_dur, step3_dur, success): |
|
| 22 |
def __init__(self, id, host, start, step1_dur, step2_dur, step3_dur, success): |
|
| 23 |
self._id = id |
|
| 23 | 24 |
self._host = host |
| 24 | 25 |
self._start = start |
| 25 | 26 |
self._step1_dur = int(step1_dur) |
| ... | ... | |
| 29 | 30 |
self._retry_1 = 0 |
| 30 | 31 |
self._retry_2 = 0 |
| 31 | 32 |
self._retry_3 = 0 |
| 33 |
self._cluster = None |
|
| 34 |
self._env = None |
|
| 35 |
|
|
| 36 |
@property |
|
| 37 |
def id(self): |
|
| 38 |
return self._id |
|
| 39 |
|
|
| 40 |
@id.setter |
|
| 41 |
def id(self, id): |
|
| 42 |
self._id = id |
|
| 32 | 43 |
|
| 33 | 44 |
@property |
| 34 | 45 |
def hostname(self): |
| ... | ... | |
| 56 | 67 |
|
| 57 | 68 |
@property |
| 58 | 69 |
def success(self): |
| 59 |
return self._success |
|
| 70 |
success = (self._step1_dur > 5 and self.step_2_duration > 10 and self.step_3_duration > 10) |
|
| 71 |
return self._success and success |
|
| 60 | 72 |
|
| 61 | 73 |
@property |
| 62 | 74 |
def num_retries_step_1(self): |
| ... | ... | |
| 88 | 100 |
self.num_retries_step_3 > 0 or |
| 89 | 101 |
self.num_retries_step_3 > 0) |
| 90 | 102 |
|
| 103 |
@property |
|
| 104 |
def cluster(self): |
|
| 105 |
return self._cluster |
|
| 106 |
|
|
| 107 |
@cluster.setter |
|
| 108 |
def cluster(self, c): |
|
| 109 |
self._cluster = c |
|
| 110 |
|
|
| 111 |
@property |
|
| 112 |
def environment(self): |
|
| 113 |
return self._env |
|
| 114 |
|
|
| 115 |
@environment.setter |
|
| 116 |
def environment(self, env): |
|
| 117 |
self._env = env |
|
| 118 |
|
|
| 119 |
|
|
| 120 |
class Deployment(list): |
|
| 121 |
""" |
|
| 122 |
Stores information about an entire deployment, |
|
| 123 |
with stats about deployment on all nodes |
|
| 124 |
""" |
|
| 125 |
|
|
| 126 |
def __init__(self, id): |
|
| 127 |
super(Deployment, self).__init__() |
|
| 128 |
self._id = id |
|
| 129 |
|
|
| 130 |
def append_stat(self, stat): |
|
| 131 |
""" |
|
| 132 |
Adds a deployment stat to this list |
|
| 133 |
:param stat: the stat to be added |
|
| 134 |
:return: nil |
|
| 135 |
""" |
|
| 136 |
self.append(stat) |
|
| 137 |
|
|
| 138 |
@property |
|
| 139 |
def id(self): |
|
| 140 |
""" |
|
| 141 |
Returns the deployment ID |
|
| 142 |
:return: the deployment ID |
|
| 143 |
""" |
|
| 144 |
return self._id |
|
| 145 |
|
|
| 146 |
@property |
|
| 147 |
def duration(self): |
|
| 148 |
""" |
|
| 149 |
Returns the time length of the longest deployment stat |
|
| 150 |
:return: the deployment duration |
|
| 151 |
""" |
|
| 152 |
last_stat = 0 |
|
| 153 |
for stat in self: |
|
| 154 |
last_stat = max(last_stat, stat.deployment_duration) |
|
| 155 |
return last_stat |
|
| 156 |
|
|
| 157 |
@property |
|
| 158 |
def success(self): |
|
| 159 |
""" |
|
| 160 |
Checks whether the deployment has been successful |
|
| 161 |
:return: False if any deployment on a machine failed |
|
| 162 |
""" |
|
| 163 |
for stat in self: |
|
| 164 |
if not stat.success: |
|
| 165 |
return False |
|
| 166 |
return True |
|
| 167 |
|
|
| 168 |
|
|
| 91 | 169 |
def parse_date(str_date): |
| 92 | 170 |
d = datetime.strptime(str_date, "%Y-%m-%d %H:%M:%S") |
| 93 | 171 |
d = TZINFO.localize(d) |
| ... | ... | |
| 99 | 177 |
help='the input request trace file') |
| 100 | 178 |
parser.add_argument('--output', dest='output', type=str, required=True,
|
| 101 | 179 |
help='the output directory') |
| 102 |
# parser.add_argument('--start-date', dest='start_date', type=parse_date, required=True,
|
|
| 103 |
# help='the start date of the log (i.e. YYYY-MM-DD HH:mm:ss)') |
|
| 104 | 180 |
parser.add_argument('--cluster', dest='cluster', type=str, required=True,
|
| 105 | 181 |
help='the name of the cluster') |
| 182 |
parser.add_argument('--plot', dest='plot', type=str, required=True, choices=['deployment'],
|
|
| 183 |
help='type of graphs that should be plotted') |
|
| 106 | 184 |
|
| 107 | 185 |
args = parser.parse_args() |
| 108 | 186 |
return args |
| ... | ... | |
| 110 | 188 |
def parse_log(log, cluster): |
| 111 | 189 |
f = open(log, 'r') |
| 112 | 190 |
|
| 113 |
deployments = []
|
|
| 191 |
stats = []
|
|
| 114 | 192 |
for line in f: |
| 115 | 193 |
fields = line.split(',')
|
| 194 |
id = fields[0] |
|
| 116 | 195 |
hostname = fields[2] |
| 117 | 196 |
|
| 118 | 197 |
if cluster in hostname: |
| ... | ... | |
| 122 | 201 |
step2_dur = int(fields[5]) |
| 123 | 202 |
step3_dur = int(fields[6]) |
| 124 | 203 |
success = "true" in fields[10] |
| 125 |
dep = DeploymentStat(hostname, start, step1_dur, step2_dur, step3_dur, success) |
|
| 204 |
env = fields[11].strip() |
|
| 205 |
dep = DeploymentStat(id, hostname, start, step1_dur, step2_dur, step3_dur, success) |
|
| 126 | 206 |
retry_1 = int(fields[7]) |
| 127 | 207 |
retry_2 = int(fields[8]) |
| 128 | 208 |
retry_3 = int(fields[9]) |
| 129 | 209 |
dep.num_retries_step_1 = retry_1 |
| 130 | 210 |
dep.num_retries_step_2 = retry_2 |
| 131 | 211 |
dep.num_retries_step_3 = retry_3 |
| 132 |
deployments.append(dep) |
|
| 212 |
dep.cluster = cluster |
|
| 213 |
dep.environment = env |
|
| 214 |
stats.append(dep) |
|
| 133 | 215 |
|
| 134 | 216 |
f.close() |
| 135 |
return deployments
|
|
| 217 |
return stats
|
|
| 136 | 218 |
|
| 137 |
def plot_deployment_time(deployments, cluster, out_dir):
|
|
| 219 |
def plot_deployment_time(stats, cluster, out_dir):
|
|
| 138 | 220 |
values = [] |
| 139 |
for d in deployments:
|
|
| 221 |
for d in stats:
|
|
| 140 | 222 |
if d.success and not d.has_retries: |
| 141 |
value = d.deployment_duration |
|
| 142 |
if d.step_1_duration > 5 and d.step_2_duration > 10 and d.step_3_duration > 10: |
|
| 143 |
values.append(value) |
|
| 223 |
values.append(d.deployment_duration) |
|
| 144 | 224 |
|
| 145 | 225 |
# title = 'Deployment on Cluster %s' % cluster |
| 146 | 226 |
title = 'Deployment on %s' % cluster |
| ... | ... | |
| 148 | 228 |
hist_graph(values, title, 'Deployment Time (seconds)', '', out_file) |
| 149 | 229 |
|
| 150 | 230 |
|
| 151 |
def plot_boot_time(deployments, cluster, out_dir): |
|
| 152 |
values = [] |
|
| 153 |
for d in deployments: |
|
| 154 |
if d.success and not d.has_retries: |
|
| 155 |
if d.step_3_duration > 30: |
|
| 156 |
values.append(d.step_3_duration) |
|
| 231 |
def group_by_deployment(stats, cluster): |
|
| 232 |
""" |
|
| 233 |
Group deployment stats by deployment. |
|
| 234 |
|
|
| 235 |
:param stats: a list of deployment stats |
|
| 236 |
:param cluster: the name of the cluster to consider |
|
| 237 |
:return: a dictionary of <id, deployment> values |
|
| 238 |
""" |
|
| 239 |
deployments = {}
|
|
| 240 |
for s in stats: |
|
| 241 |
if s.cluster != cluster: |
|
| 242 |
continue |
|
| 243 |
|
|
| 244 |
if s.id in deployments: |
|
| 245 |
dep = deployments[s.id] |
|
| 246 |
else: |
|
| 247 |
dep = Deployment(s.id) |
|
| 248 |
deployments[s.id] = dep |
|
| 249 |
dep.append_stat(s) |
|
| 250 |
return deployments |
|
| 251 |
|
|
| 157 | 252 |
|
| 158 |
title = 'Machine Boot Time on Cluster %s' % (cluster) |
|
| 159 |
out_file = out_dir + ("/boot_time_%s.pdf" % (cluster))
|
|
| 160 |
hist_graph(values, title, 'Boot Time (seconds)', '', out_file) |
|
| 253 |
def plot_deployment_size(deployments, cluster): |
|
| 254 |
sizes = {}
|
|
| 255 |
for d in deployments.itervalues(): |
|
| 256 |
size = len(d) |
|
| 257 |
if size not in sizes: |
|
| 258 |
sizes[size] = [] |
|
| 161 | 259 |
|
| 260 |
if size > 0 and d.success: |
|
| 261 |
sizes[size].append(d.duration) |
|
| 262 |
|
|
| 263 |
fig, ax = plt.subplots() |
|
| 264 |
x = [] |
|
| 265 |
means = [] |
|
| 266 |
stds = [] |
|
| 267 |
for size, value in sizes.iteritems(): |
|
| 268 |
if len(value) > 0: |
|
| 269 |
x.append(size) |
|
| 270 |
means.append(np.mean(value)) |
|
| 271 |
stds.append(np.std(value)) |
|
| 272 |
|
|
| 273 |
ax.set_xlim(0, np.max(x) + 1) |
|
| 274 |
ax.set_ylabel('Deployment Duration (s)')
|
|
| 275 |
ax.set_xlabel('Deployment Size (\# Machines)')
|
|
| 276 |
|
|
| 277 |
plt.errorbar(x, means, yerr=stds) |
|
| 278 |
plt.show() |
|
| 162 | 279 |
|
| 163 | 280 |
def hist_graph(values, title, x_label, y_label, out_file): |
| 164 | 281 |
ar1 = np.array(values) |
| ... | ... | |
| 211 | 328 |
#plt.show() |
| 212 | 329 |
plt.savefig(out_file) |
| 213 | 330 |
|
| 331 |
|
|
| 214 | 332 |
def plot_graphs(): |
| 215 | 333 |
opts = parse_kdeploy_opt() |
| 216 |
deployments = parse_log(opts.input, opts.cluster) |
|
| 217 |
plot_deployment_time(deployments, opts.cluster, opts.output) |
|
| 218 |
# plot_boot_time(deployments, opts.cluster, opts.output) |
|
| 334 |
if opts.plot == 'deployment': |
|
| 335 |
print 'Printing deployment graphs...' |
|
| 336 |
stats = parse_log(opts.input, opts.cluster) |
|
| 337 |
# plot_deployment_time(stats, opts.cluster, opts.output) |
|
| 338 |
deployments = group_by_deployment(stats, opts.cluster) |
|
| 339 |
plot_deployment_size(deployments, opts.cluster) |
|
Formats disponibles : Unified diff