skylight 0.3.0 → 0.3.1
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +4 -4
- data/CHANGELOG.md +32 -0
- data/lib/skylight.rb +14 -8
- data/lib/skylight/api.rb +0 -2
- data/lib/skylight/version.rb +1 -1
- metadata +3 -4
- data/too-many-sockets.md +0 -62
checksums.yaml
CHANGED
@@ -1,7 +1,7 @@
|
|
1
1
|
---
|
2
2
|
SHA1:
|
3
|
-
metadata.gz:
|
4
|
-
data.tar.gz:
|
3
|
+
metadata.gz: 4005d9f17790586799b47b937adeaf32049239cb
|
4
|
+
data.tar.gz: b58afb77fd648bfbdc904a305161fb57a47ceeb1
|
5
5
|
SHA512:
|
6
|
-
metadata.gz:
|
7
|
-
data.tar.gz:
|
6
|
+
metadata.gz: e7c5939f477c3902f1684b20fb5030aef6bdc7554dd395589792dcbd6421224b75a60a9fd0e8d4292296afede7c3e582e8cf399af866e67ad8d2f7b855ccbfd1
|
7
|
+
data.tar.gz: 9c5a136bc94db3c8bc51c80ec4ae191a23d6cb5e9eded4dc0e5bdcb1a60e5e1dd006bad9965bb9c9402195b8684849cb4783ec4160b705bbab9800ec2d586789
|
data/CHANGELOG.md
CHANGED
@@ -1,3 +1,35 @@
|
|
1
|
+
## 0.3.1 (March 8, 2014)
|
2
|
+
|
3
|
+
* Fix requires to allow CLI to function without native extension.
|
4
|
+
|
5
|
+
## 0.3.0 (February 28, 2014)
|
6
|
+
|
7
|
+
* Native Rust agent
|
8
|
+
* Send exceptions occurring during HTTP requests to the client.
|
9
|
+
* Warn users when skylight is potentially disabled incorrectly.
|
10
|
+
* Update SQL Lexer to 0.0.6
|
11
|
+
* Log the backtraces of unhandled exceptions
|
12
|
+
* Add support for disabling GC tracking
|
13
|
+
* Add support for disabling agent
|
14
|
+
|
15
|
+
## 0.2.7 (February 26, 2014)
|
16
|
+
|
17
|
+
* Disable annotations to reduce memory load.
|
18
|
+
|
19
|
+
## 0.2.6 (February 25, 2014)
|
20
|
+
|
21
|
+
* `inspect` even whitelisted payload props
|
22
|
+
* Ignore Errno::EINTR for 'ps' call
|
23
|
+
|
24
|
+
## 0.2.5 (February 21, 2014)
|
25
|
+
|
26
|
+
* Revert "Update SqlLexer to 0.0.4"
|
27
|
+
|
28
|
+
## 0.2.4 (February 20, 2014)
|
29
|
+
|
30
|
+
* Whitelist process action annotation keys.
|
31
|
+
* Update SqlLexer to 0.0.4
|
32
|
+
|
1
33
|
## 0.2.3 (December 20, 2013)
|
2
34
|
|
3
35
|
* Fix SQL lexing for comments, arrays, double-colon casting, and multiple queries
|
data/lib/skylight.rb
CHANGED
@@ -13,10 +13,22 @@ rescue LoadError
|
|
13
13
|
raise if ENV.key?("SKYLIGHT_REQUIRED")
|
14
14
|
end
|
15
15
|
|
16
|
+
module Skylight
|
17
|
+
TRACE_ENV_KEY = 'SKYLIGHT_ENABLE_TRACE_LOGS'.freeze
|
18
|
+
|
19
|
+
autoload :Api, 'skylight/api'
|
20
|
+
autoload :CLI, 'skylight/cli'
|
21
|
+
autoload :Config, 'skylight/config'
|
22
|
+
|
23
|
+
module Util
|
24
|
+
autoload :Logging, 'skylight/util/logging'
|
25
|
+
autoload :HTTP, 'skylight/util/http'
|
26
|
+
end
|
27
|
+
end
|
28
|
+
|
16
29
|
if has_native_ext
|
17
30
|
|
18
31
|
module Skylight
|
19
|
-
TRACE_ENV_KEY = 'SKYLIGHT_ENABLE_TRACE_LOGS'.freeze
|
20
32
|
STANDALONE_ENV_KEY = 'SKYLIGHT_STANDALONE'.freeze
|
21
33
|
STANDALONE_ENV_VAL = 'server'.freeze
|
22
34
|
|
@@ -32,7 +44,6 @@ module Skylight
|
|
32
44
|
require 'skylight/vm/gc'
|
33
45
|
end
|
34
46
|
|
35
|
-
autoload :Config, 'skylight/config'
|
36
47
|
autoload :GC, 'skylight/gc'
|
37
48
|
autoload :Helpers, 'skylight/helpers'
|
38
49
|
autoload :Instrumenter, 'skylight/instrumenter'
|
@@ -175,9 +186,4 @@ module Skylight
|
|
175
186
|
end
|
176
187
|
end
|
177
188
|
|
178
|
-
end
|
179
|
-
|
180
|
-
module Skylight
|
181
|
-
autoload :Api, 'skylight/api'
|
182
|
-
autoload :CLI, 'skylight/cli'
|
183
|
-
end
|
189
|
+
end
|
data/lib/skylight/api.rb
CHANGED
data/lib/skylight/version.rb
CHANGED
metadata
CHANGED
@@ -1,14 +1,14 @@
|
|
1
1
|
--- !ruby/object:Gem::Specification
|
2
2
|
name: skylight
|
3
3
|
version: !ruby/object:Gem::Version
|
4
|
-
version: 0.3.
|
4
|
+
version: 0.3.1
|
5
5
|
platform: ruby
|
6
6
|
authors:
|
7
7
|
- Tilde, Inc.
|
8
8
|
autorequire:
|
9
9
|
bindir: bin
|
10
10
|
cert_chain: []
|
11
|
-
date: 2014-
|
11
|
+
date: 2014-03-09 00:00:00.000000000 Z
|
12
12
|
dependencies:
|
13
13
|
- !ruby/object:Gem::Dependency
|
14
14
|
name: activesupport
|
@@ -139,7 +139,6 @@ files:
|
|
139
139
|
- lib/sql_lexer.rb
|
140
140
|
- lib/sql_lexer/lexer.rb
|
141
141
|
- lib/sql_lexer/version.rb
|
142
|
-
- too-many-sockets.md
|
143
142
|
homepage: http://www.skylight.io
|
144
143
|
licenses: []
|
145
144
|
metadata: {}
|
@@ -159,7 +158,7 @@ required_rubygems_version: !ruby/object:Gem::Requirement
|
|
159
158
|
version: '0'
|
160
159
|
requirements: []
|
161
160
|
rubyforge_project:
|
162
|
-
rubygems_version: 2.2.
|
161
|
+
rubygems_version: 2.2.1
|
163
162
|
signing_key:
|
164
163
|
specification_version: 4
|
165
164
|
summary: Skylight is a ruby application monitoring tool. Currently in closed beta.
|
data/too-many-sockets.md
DELETED
@@ -1,62 +0,0 @@
|
|
1
|
-
We received a notification (Zendesk #280) from a customer that they were unable to SSH
|
2
|
-
into their box, and that their service provider discovered that Skylight
|
3
|
-
was triggering socket buffer limits.
|
4
|
-
|
5
|
-
The customer immediately disabled Skylight, which fixed the bug for him.
|
6
|
-
|
7
|
-
Customer Information:
|
8
|
-
|
9
|
-
* Using Skylight 0.2.3
|
10
|
-
* Ruby 2.0.0p353
|
11
|
-
* Also had New Relic installed (Bug #46)
|
12
|
-
* Recently upgraded to Rails 4
|
13
|
-
* Passenger Enterprise with a max-pool size of 12 and zero-downtime deploys
|
14
|
-
* He had a maxsockbuf of 32MB
|
15
|
-
|
16
|
-
The customer also reported 134 open sockets in the server process.
|
17
|
-
|
18
|
-
Separately, we discovered that New Relic was making our payloads
|
19
|
-
extremely large.
|
20
|
-
|
21
|
-
Our working hypothesis was that the large New Relic payloads were
|
22
|
-
triggering the max buffer condition, which was then cascading into more
|
23
|
-
failures.
|
24
|
-
|
25
|
-
Upon further investigation, we discovered a few things:
|
26
|
-
|
27
|
-
* If the client hits a kernel buffer limit (via EWOULDBLOCK that lasts
|
28
|
-
more than 5s), it closes down its socket
|
29
|
-
* The server expects to recover from this situation by getting the
|
30
|
-
client socket in its read list from IO.select. It would then try to
|
31
|
-
read from it, get an EOF, and close the socket.
|
32
|
-
* If the server does not get the client socket in its read list for some
|
33
|
-
(unknown) reason, this would result in an ever-growing list of sockets
|
34
|
-
on the server side.
|
35
|
-
|
36
|
-
We are not sure why the sockets would not appear in the read list, but
|
37
|
-
we hypothesize that when the OpenVZ limit is reached, the kernel no
|
38
|
-
longer includes the socket in the read list, causing the server to never
|
39
|
-
close the socket.
|
40
|
-
|
41
|
-
Mitigation strategies:
|
42
|
-
|
43
|
-
* The New Relic fix (#46) should reduce the likelihood of encountering
|
44
|
-
this in the first place
|
45
|
-
* We want to add a Hello heartbeat from the client. If the server
|
46
|
-
doesn't receive a message every 1m (or 2m, TBD), it will close the
|
47
|
-
socket even if it's not in the read list.
|
48
|
-
|
49
|
-
We are worried that if the server process ever gets stuck, this
|
50
|
-
condition can occur. We are also considering a server heartbeat back to
|
51
|
-
the clients, so they can take corrective action if the agent gets stuck.
|
52
|
-
|
53
|
-
Some mitigations for this situation:
|
54
|
-
|
55
|
-
* Kill the server process when a respawn is necessary. This will ensure
|
56
|
-
that a stuck agent process doesn't duplicate into N servers.
|
57
|
-
* Before connecting to the server socket, pre-check the server process
|
58
|
-
with kill -0.
|
59
|
-
|
60
|
-
For next time:
|
61
|
-
* Request skylight.log for easier debugging
|
62
|
-
* Form a best practices for customer bug reporting on doc site
|