-
Notifications
You must be signed in to change notification settings - Fork 33
/
main.py
executable file
·276 lines (240 loc) · 19.7 KB
/
main.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
#!/usr/bin/env python3
import os
import time
from helpers import INTERVAL_TIME, PROMETHEUS_URL, DRY_RUN, VERBOSE, get_settings_for_prometheus_metrics, is_integer_or_float, print_human_readable_volume_dict
from helpers import convert_bytes_to_storage, scale_up_pvc, testIfPrometheusIsAccessible, describe_all_pvcs, send_kubernetes_event
from helpers import fetch_pvcs_from_prometheus, printHeaderAndConfiguration, calculateBytesToScaleTo, GracefulKiller, cache
from prometheus_client import start_http_server, Summary, Gauge, Counter, Info
import slack
import sys, traceback
# Initialize our Prometheus metrics (counters)
PROMETHEUS_METRICS = {}
PROMETHEUS_METRICS['resize_evaluated'] = Counter('volume_autoscaler_resize_evaluated', 'Counter which is increased every time we evaluate resizing PVCs')
PROMETHEUS_METRICS['resize_attempted'] = Counter('volume_autoscaler_resize_attempted', 'Counter which is increased every time we attempt to resize')
PROMETHEUS_METRICS['resize_successful'] = Counter('volume_autoscaler_resize_successful', 'Counter which is increased every time we successfully resize')
PROMETHEUS_METRICS['resize_failure'] = Counter('volume_autoscaler_resize_failure', 'Counter which is increased every time we fail to resize')
# Initialize our Prometheus metrics (gauges)
PROMETHEUS_METRICS['num_valid_pvcs'] = Gauge('volume_autoscaler_num_valid_pvcs', 'Gauge with the number of valid PVCs detected which we found to consider for scaling')
PROMETHEUS_METRICS['num_valid_pvcs'].set(0)
PROMETHEUS_METRICS['num_pvcs_above_threshold'] = Gauge('volume_autoscaler_num_pvcs_above_threshold', 'Gauge with the number of PVCs detected above the desired percentage threshold')
PROMETHEUS_METRICS['num_pvcs_above_threshold'].set(0)
PROMETHEUS_METRICS['num_pvcs_below_threshold'] = Gauge('volume_autoscaler_num_pvcs_below_threshold', 'Gauge with the number of PVCs detected below the desired percentage threshold')
PROMETHEUS_METRICS['num_pvcs_below_threshold'].set(0)
# Initialize our Prometheus metrics (info/settings)
PROMETHEUS_METRICS['info'] = Info('volume_autoscaler_release', 'Release/version information about this volume autoscaler service')
PROMETHEUS_METRICS['info'].info({'version': '1.0.7'})
PROMETHEUS_METRICS['settings'] = Info('volume_autoscaler_settings', 'Settings currently used in this service')
PROMETHEUS_METRICS['settings'].info(get_settings_for_prometheus_metrics())
# Other globals
MAIN_LOOP_TIME = 1
# Entry point and main application loop
if __name__ == "__main__":
# Test if our prometheus URL works before continuing
testIfPrometheusIsAccessible(PROMETHEUS_URL)
# Startup our prometheus metrics endpoint
start_http_server(8000)
# TODO: Test k8s access, or just test on-the-fly below?
# Reporting our configuration to the end-user
printHeaderAndConfiguration()
# Setup our graceful handling of kubernetes signals
killer = GracefulKiller()
last_run = 0
# Our main run loop, now using a signal handler to handle kubernetes signals gracefully (not mid-loop)
while not killer.kill_now:
# If it's not our interval time yet, only run once every INTERVAL_TIME seconds. This extra bit helps us handle signals gracefully quicker
if int(time.time()) - last_run <= INTERVAL_TIME:
time.sleep(MAIN_LOOP_TIME)
continue
last_run = int(time.time())
# In every loop, fetch all our pvcs state from Kubernetes
try:
PROMETHEUS_METRICS['resize_evaluated'].inc()
pvcs_in_kubernetes = describe_all_pvcs(simple=True)
except Exception:
print("Exception while trying to describe all PVCs")
traceback.print_exc()
time.sleep(MAIN_LOOP_TIME)
continue
# Fetch our volume usage from Prometheus
try:
pvcs_in_prometheus = fetch_pvcs_from_prometheus(url=PROMETHEUS_URL)
print("Querying and found {} valid PVCs to assess in prometheus".format(len(pvcs_in_prometheus)))
PROMETHEUS_METRICS['num_valid_pvcs'].set(len(pvcs_in_prometheus))
except Exception:
print("Exception while trying to fetch PVC metrics from prometheus")
traceback.print_exc()
time.sleep(MAIN_LOOP_TIME)
continue
# Iterate through every item and handle it accordingly
PROMETHEUS_METRICS['num_pvcs_above_threshold'].set(0) # Reset these each loop
PROMETHEUS_METRICS['num_pvcs_below_threshold'].set(0) # Reset these each loop
for item in pvcs_in_prometheus:
try:
volume_name = str(item['metric']['persistentvolumeclaim'])
volume_namespace = str(item['metric']['namespace'])
volume_description = "{}.{}".format(item['metric']['namespace'], item['metric']['persistentvolumeclaim'])
volume_used_percent = int(item['value'][1])
# Precursor check to ensure we have info for this pvc in kubernetes object
if volume_description not in pvcs_in_kubernetes:
print("ERROR: The volume {} was not found in Kubernetes but had metrics in Prometheus. This may be an old volume, was just deleted, or some random jitter is occurring. If this continues to occur, please report an bug. You might also be using an older version of Prometheus, please make sure you're using v2.30.0 or newer before reporting a bug for this.".format(volume_description))
continue
pvcs_in_kubernetes[volume_description]['volume_used_percent'] = volume_used_percent
try:
volume_used_inode_percent = int(item['value_inodes'])
except:
volume_used_inode_percent = -1
pvcs_in_kubernetes[volume_description]['volume_used_inode_percent'] = volume_used_inode_percent
if VERBOSE:
print(" VERBOSE DETAILS:")
print("-------------------------------------------------------------------------------------------------------------")
print_human_readable_volume_dict(pvcs_in_kubernetes[volume_description])
print("-------------------------------------------------------------------------------------------------------------")
print("Volume {} has {}% disk space used of the {} available".format(volume_description,volume_used_percent,pvcs_in_kubernetes[volume_description]['volume_size_status']))
if volume_used_inode_percent > -1:
print("Volume {} has {}% inodes used".format(volume_description,volume_used_inode_percent))
# Check if we are NOT in an alert condition
if volume_used_percent < pvcs_in_kubernetes[volume_description]['scale_above_percent'] and volume_used_inode_percent < pvcs_in_kubernetes[volume_description]['scale_above_percent']:
PROMETHEUS_METRICS['num_pvcs_below_threshold'].inc()
cache.unset(volume_description)
if VERBOSE:
print(" and is not above {}% used".format(pvcs_in_kubernetes[volume_description]['scale_above_percent']))
if volume_used_inode_percent > -1:
print(" and is not above {}% inodes used".format(pvcs_in_kubernetes[volume_description]['scale_above_percent']))
if VERBOSE:
print("=============================================================================================================")
continue
else:
PROMETHEUS_METRICS['num_pvcs_above_threshold'].inc()
# If we are in alert condition, record this in our simple in-memory counter
if cache.get(volume_description):
cache.set(volume_description, cache.get(volume_description) + 1)
else:
cache.set(volume_description, 1)
# Incase we aren't verbose, and didn't print this above, now that we're in alert we will print this
if not VERBOSE:
print("Volume {} is {}% in-use of the {} available".format(volume_description,volume_used_percent,pvcs_in_kubernetes[volume_description]['volume_size_status']))
print("Volume {} is {}% inode in-use".format(volume_description,volume_used_inode_percent))
# Print the alert status and reason
if volume_used_percent >= pvcs_in_kubernetes[volume_description]['scale_above_percent']:
print(" BECAUSE it has space used above {}%".format(pvcs_in_kubernetes[volume_description]['scale_above_percent']))
elif volume_used_inode_percent >= pvcs_in_kubernetes[volume_description]['scale_above_percent']:
print(" BECAUSE it has inodes used above {}%".format(pvcs_in_kubernetes[volume_description]['scale_above_percent']))
print(" ALERT has been for {} period(s) which needs to at least {} period(s) to scale".format(cache.get(volume_description), pvcs_in_kubernetes[volume_description]['scale_after_intervals']))
# Check if we are NOT in a possible scale condition
if cache.get(volume_description) < pvcs_in_kubernetes[volume_description]['scale_after_intervals']:
print(" BUT need to wait for {} intervals in alert before considering to scale".format( pvcs_in_kubernetes[volume_description]['scale_after_intervals'] ))
print(" FYI this has desired_size {} and current size {}".format( convert_bytes_to_storage(pvcs_in_kubernetes[volume_description]['volume_size_spec_bytes']), convert_bytes_to_storage(pvcs_in_kubernetes[volume_description]['volume_size_status_bytes'])))
print("=============================================================================================================")
continue
# If we are in a possible scale condition, check if we recently scaled it and handle accordingly
if pvcs_in_kubernetes[volume_description]['last_resized_at'] + pvcs_in_kubernetes[volume_description]['scale_cooldown_time'] >= int(time.mktime(time.gmtime())):
print(" BUT need to wait {} seconds to scale since the last scale time {} seconds ago".format( abs(pvcs_in_kubernetes[volume_description]['last_resized_at'] + pvcs_in_kubernetes[volume_description]['scale_cooldown_time']) - int(time.mktime(time.gmtime())), abs(pvcs_in_kubernetes[volume_description]['last_resized_at'] - int(time.mktime(time.gmtime()))) ))
print("=============================================================================================================")
continue
# If we reach this far then we will be scaling the disk, all preconditions were passed from above
if pvcs_in_kubernetes[volume_description]['last_resized_at'] == 0:
print(" AND we need to scale it immediately, it has never been scaled previously")
else:
print(" AND we need to scale it immediately, it last scaled {} seconds ago".format( abs((pvcs_in_kubernetes[volume_description]['last_resized_at'] + pvcs_in_kubernetes[volume_description]['scale_cooldown_time']) - int(time.mktime(time.gmtime()))) ))
# Calculate how many bytes to resize to based on the parameters provided globally and per-this pv annotations
resize_to_bytes = calculateBytesToScaleTo(
original_size = pvcs_in_kubernetes[volume_description]['volume_size_status_bytes'],
scale_up_percent = pvcs_in_kubernetes[volume_description]['scale_up_percent'],
min_increment = pvcs_in_kubernetes[volume_description]['scale_up_min_increment'],
max_increment = pvcs_in_kubernetes[volume_description]['scale_up_max_increment'],
maximum_size = pvcs_in_kubernetes[volume_description]['scale_up_max_size'],
)
# TODO: Check here if storage class has the ALLOWVOLUMEEXPANSION flag set to true, read the SC from pvcs_in_kubernetes[volume_description]['storage_class'] ?
# If our resize bytes failed for some reason, eg putting invalid data into the annotations on the PV
if resize_to_bytes == False:
print("-------------------------------------------------------------------------------------------------------------")
print(" Error/Exception while trying to determine what to resize to, volume causing failure:")
print("-------------------------------------------------------------------------------------------------------------")
print(pvcs_in_kubernetes[volume_description])
print("=============================================================================================================")
continue
# If our resize bytes is less than our original size (because the user set the max-bytes to something too low)
if resize_to_bytes < pvcs_in_kubernetes[volume_description]['volume_size_status_bytes']:
print("-------------------------------------------------------------------------------------------------------------")
print(" Error/Exception while trying to scale this up. Is it possible your maximum SCALE_UP_MAX_SIZE is too small?")
print("-------------------------------------------------------------------------------------------------------------")
print(" Maximum Size: {} ({})".format(pvcs_in_kubernetes[volume_description]['scale_up_max_size'], convert_bytes_to_storage(pvcs_in_kubernetes[volume_description]['scale_up_max_size'])))
print(" Original Size: {} ({})".format(pvcs_in_kubernetes[volume_description]['volume_size_status_bytes'], convert_bytes_to_storage(pvcs_in_kubernetes[volume_description]['volume_size_status_bytes'])))
print(" Resize To: {} ({})".format(resize_to_bytes, convert_bytes_to_storage(resize_to_bytes)))
print("-------------------------------------------------------------------------------------------------------------")
print(" Volume causing failure:")
print_human_readable_volume_dict(pvcs_in_kubernetes[volume_description])
print("=============================================================================================================")
continue
# Check if we are already at the max volume size (either globally, or this-volume specific)
if resize_to_bytes == pvcs_in_kubernetes[volume_description]['volume_size_status_bytes']:
print(" SKIPPING scaling this because we are at the maximum size of {}".format(convert_bytes_to_storage(pvcs_in_kubernetes[volume_description]['scale_up_max_size'])))
print("=============================================================================================================")
continue
# Check if we set on this PV we want to ignore the volume autoscaler
if pvcs_in_kubernetes[volume_description]['ignore']:
print(" IGNORING scaling this because the ignore annotation was set to true")
print("=============================================================================================================")
continue
# Lets debounce this incase we did this resize last interval(s)
if cache.get(f"{volume_description}-has-been-resized"):
print(" DEBOUNCING and skipping this scaling, we resized within recent intervals")
print("=============================================================================================================")
continue
# Check if we are DRY-RUN-ing and won't do anything
if DRY_RUN:
print(" DRY RUN was set, but we would have resized this disk from {} to {}".format(convert_bytes_to_storage(pvcs_in_kubernetes[volume_description]['volume_size_status_bytes']), convert_bytes_to_storage(resize_to_bytes)))
print("=============================================================================================================")
continue
# If we aren't dry-run, lets resize
PROMETHEUS_METRICS['resize_attempted'].inc()
print(" RESIZING disk from {} to {}".format(convert_bytes_to_storage(pvcs_in_kubernetes[volume_description]['volume_size_status_bytes']), convert_bytes_to_storage(resize_to_bytes)))
status_output = "to scale up `{}` by `{}%` from `{}` to `{}`, it was using more than `{}%` disk or inode space over the last `{} seconds`".format(
volume_description,
pvcs_in_kubernetes[volume_description]['scale_up_percent'],
convert_bytes_to_storage(pvcs_in_kubernetes[volume_description]['volume_size_status_bytes']),
convert_bytes_to_storage(resize_to_bytes),
pvcs_in_kubernetes[volume_description]['scale_above_percent'],
cache.get(volume_description) * INTERVAL_TIME
)
# Send event that we're starting to request a resize
send_kubernetes_event(
name=volume_name, namespace=volume_namespace, reason="VolumeResizeRequested",
message="Requesting {}".format(status_output)
)
if scale_up_pvc(volume_namespace, volume_name, resize_to_bytes):
PROMETHEUS_METRICS['resize_successful'].inc()
# Save this to cache for debouncing
cache.set(f"{volume_description}-has-been-resized", True)
# Print success to console
status_output = "Successfully requested {}".format(status_output)
print(status_output)
# Intentionally skipping sending an event to Kubernetes on success, the above event is enough for now until we detect if resize succeeded
# Print success to Slack
if slack.SLACK_WEBHOOK_URL and len(slack.SLACK_WEBHOOK_URL) > 0:
print(f"Sending slack message to {slack.SLACK_CHANNEL}")
slack.send(status_output)
else:
PROMETHEUS_METRICS['resize_failure'].inc()
# Print failure to console
status_output = "FAILED requesting {}".format(status_output)
print(status_output)
# Print failure to Kubernetes Events
send_kubernetes_event(
name=volume_name, namespace=volume_namespace, reason="VolumeResizeRequestFailed",
message=status_output, type="Warning"
)
# Print failure to Slack
if slack.SLACK_WEBHOOK_URL and len(slack.SLACK_WEBHOOK_URL) > 0:
print(f"Sending slack message to {slack.SLACK_CHANNEL}")
slack.send(status_output, severity="error")
except Exception:
print("Exception caught while trying to process record")
print(item)
traceback.print_exc()
if VERBOSE:
print("=============================================================================================================")
# Wait until our next interval
time.sleep(MAIN_LOOP_TIME)
print("We were sent a signal handler to kill, exited gracefully")
exit(0)