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
|
---
stage: none
group: unassigned
info: Any user with at least the Maintainer role can merge updates to this content. For details, see https://docs.gitlab.com/ee/development/development_processes.html#development-guidelines-review.
---
# Generating chaos in a test GitLab instance
<!-- vale gitlab_base.Spelling = NO -->
As [Werner Vogels](https://twitter.com/Werner), the CTO at Amazon Web Services, famously put it, **Everything fails, all the time**.
<!-- vale gitlab_base.Spelling = NO -->
As a developer, it's as important to consider the failure modes in which your software may operate as much as typical operation. Doing so can mean the difference between a minor hiccup leading to a scattering of `500` errors experienced by a tiny fraction of users, and a full site outage that affects all users for an extended period.
To paraphrase [Tolstoy](https://en.wikipedia.org/wiki/Anna_Karenina_principle), _all happy servers are alike, but all failing servers are failing in their own way_. Luckily, there are ways we can attempt to simulate these failure modes, and the chaos endpoints are tools for assisting in this process.
Currently, there are four endpoints for simulating the following conditions:
- Slow requests.
- CPU-bound requests.
- Memory leaks.
- Unexpected process crashes.
## Enabling chaos endpoints
For obvious reasons, these endpoints are not enabled by default on `production`.
They are enabled by default on **development** environments.
WARNING:
It is required that you secure access to the chaos endpoints using a secret token.
You should not enable them in production unless you absolutely know what you're doing.
A secret token can be set through the `GITLAB_CHAOS_SECRET` environment variable.
For example, when using the [GDK](https://gitlab.com/gitlab-org/gitlab-development-kit)
this can be done with the following command:
```shell
GITLAB_CHAOS_SECRET=secret gdk start
```
Replace `secret` with your own secret token.
## Invoking chaos
After you have enabled the chaos endpoints and restarted the application, you can start testing using the endpoints.
By default, when invoking a chaos endpoint, the web worker process which receives the request handles it. This means, for example, that if the Kill
operation is invoked, the Puma worker process handling the request is killed. To test these operations in Sidekiq, the `async` parameter on
each endpoint can be set to `true`. This runs the chaos process in a Sidekiq worker.
## Memory leaks
To simulate a memory leak in your application, use the `/-/chaos/leakmem` endpoint.
The memory is not retained after the request finishes. After the request has completed, the Ruby garbage collector attempts to recover the memory.
```plaintext
GET /-/chaos/leakmem
GET /-/chaos/leakmem?memory_mb=1024
GET /-/chaos/leakmem?memory_mb=1024&duration_s=50
GET /-/chaos/leakmem?memory_mb=1024&duration_s=50&async=true
```
| Attribute | Type | Required | Description |
| ------------ | ------- | -------- | ------------------------------------------------------------------------------------ |
| `memory_mb` | integer | no | How much memory, in MB, should be leaked. Defaults to 100 MB. |
| `duration_s` | integer | no | Minimum duration_s, in seconds, that the memory should be retained. Defaults to 30 s. |
| `async` | boolean | no | Set to true to leak memory in a Sidekiq background worker process |
```shell
curl "http://localhost:3000/-/chaos/leakmem?memory_mb=1024&duration_s=10" \
--header 'X-Chaos-Secret: secret'
curl "http://localhost:3000/-/chaos/leakmem?memory_mb=1024&duration_s=10&token=secret"
```
## CPU spin
This endpoint attempts to fully use a single core, at 100%, for the given period.
Depending on your rack server setup, your request may timeout after a predetermined period (typically 60 seconds).
```plaintext
GET /-/chaos/cpu_spin
GET /-/chaos/cpu_spin?duration_s=50
GET /-/chaos/cpu_spin?duration_s=50&async=true
```
| Attribute | Type | Required | Description |
| ------------ | ------- | -------- | --------------------------------------------------------------------- |
| `duration_s` | integer | no | Duration, in seconds, that the core is used. Defaults to 30 s |
| `async` | boolean | no | Set to true to consume CPU in a Sidekiq background worker process |
```shell
curl "http://localhost:3000/-/chaos/cpu_spin?duration_s=60" \
--header 'X-Chaos-Secret: secret'
curl "http://localhost:3000/-/chaos/cpu_spin?duration_s=60&token=secret"
```
## DB spin
This endpoint attempts to fully use a single core, and interleave it with DB request, for the given period.
This endpoint can be used to model yielding execution to another threads when running concurrently.
Depending on your rack server setup, your request may timeout after a predetermined period (typically 60 seconds).
```plaintext
GET /-/chaos/db_spin
GET /-/chaos/db_spin?duration_s=50
GET /-/chaos/db_spin?duration_s=50&async=true
```
| Attribute | Type | Required | Description |
| ------------ | ------- | -------- | --------------------------------------------------------------------------- |
| `interval_s` | float | no | Interval, in seconds, for every DB request. Defaults to 1 s |
| `duration_s` | integer | no | Duration, in seconds, that the core is used. Defaults to 30 s |
| `async` | boolean | no | Set to true to perform the operation in a Sidekiq background worker process |
```shell
curl "http://localhost:3000/-/chaos/db_spin?interval_s=1&duration_s=60" \
--header 'X-Chaos-Secret: secret'
curl "http://localhost:3000/-/chaos/db_spin?interval_s=1&duration_s=60&token=secret"
```
## Sleep
This endpoint is similar to the CPU Spin endpoint but simulates off-processor activity, such as network calls to backend services. It sleeps for a given `duration_s`.
As with the CPU Spin endpoint, this may lead to your request timing out if `duration_s` exceeds the configured limit.
```plaintext
GET /-/chaos/sleep
GET /-/chaos/sleep?duration_s=50
GET /-/chaos/sleep?duration_s=50&async=true
```
| Attribute | Type | Required | Description |
| ------------ | ------- | -------- | ---------------------------------------------------------------------- |
| `duration_s` | integer | no | Duration, in seconds, that the request sleeps for. Defaults to 30 s |
| `async` | boolean | no | Set to true to sleep in a Sidekiq background worker process |
```shell
curl "http://localhost:3000/-/chaos/sleep?duration_s=60" \
--header 'X-Chaos-Secret: secret'
curl "http://localhost:3000/-/chaos/sleep?duration_s=60&token=secret"
```
## Kill
This endpoint simulates the unexpected death of a worker process using the `KILL` signal.
Because this endpoint uses the `KILL` signal, the process isn't given an
opportunity to clean up or shut down.
```plaintext
GET /-/chaos/kill
GET /-/chaos/kill?async=true
```
| Attribute | Type | Required | Description |
| ------------ | ------- | -------- | ---------------------------------------------------------------------- |
| `async` | boolean | no | Set to true to signal a Sidekiq background worker process |
```shell
curl "http://localhost:3000/-/chaos/kill" --header 'X-Chaos-Secret: secret'
curl "http://localhost:3000/-/chaos/kill?token=secret"
```
## Quit
This endpoint simulates the unexpected death of a worker process using the `QUIT` signal.
Unlike `KILL`, the `QUIT` signal also attempts to write a core dump.
See [core(5)](https://man7.org/linux/man-pages/man5/core.5.html) for more information.
```plaintext
GET /-/chaos/quit
GET /-/chaos/quit?async=true
```
| Attribute | Type | Required | Description |
| ------------ | ------- | -------- | ---------------------------------------------------------------------- |
| `async` | boolean | no | Set to true to signal a Sidekiq background worker process |
```shell
curl "http://localhost:3000/-/chaos/quit" --header 'X-Chaos-Secret: secret'
curl "http://localhost:3000/-/chaos/quit?token=secret"
```
## Run garbage collector
This endpoint triggers a GC run on the worker handling the request and returns its worker ID
plus GC stats as JSON. This is mostly useful when running Puma in standalone mode, since
otherwise the worker handling the request cannot be known upfront.
Endpoint:
```plaintext
POST /-/chaos/gc
```
Example request:
```shell
curl --request POST "http://localhost:3000/-/chaos/gc" \
--header 'X-Chaos-Secret: secret'
curl --request POST "http://localhost:3000/-/chaos/gc?token=secret"
```
Example response:
```json
{
"worker_id": "puma_1",
"gc_stat": {
"count": 94,
"heap_allocated_pages": 9077,
"heap_sorted_length": 9077,
"heap_allocatable_pages": 0,
"heap_available_slots": 3699720,
"heap_live_slots": 2827510,
"heap_free_slots": 872210,
"heap_final_slots": 0,
"heap_marked_slots": 2827509,
"heap_eden_pages": 9077,
"heap_tomb_pages": 0,
"total_allocated_pages": 9077,
"total_freed_pages": 0,
"total_allocated_objects": 14229357,
"total_freed_objects": 11401847,
"malloc_increase_bytes": 8192,
"malloc_increase_bytes_limit": 30949538,
"minor_gc_count": 71,
"major_gc_count": 23,
"compact_count": 0,
"remembered_wb_unprotected_objects": 41685,
"remembered_wb_unprotected_objects_limit": 83370,
"old_objects": 2617806,
"old_objects_limit": 5235612,
"oldmalloc_increase_bytes": 8192,
"oldmalloc_increase_bytes_limit": 122713697
}
}
```
|
