forked from defunkt/unicorn
-
Notifications
You must be signed in to change notification settings - Fork 0
/
KNOWN_ISSUES
79 lines (62 loc) · 3.79 KB
/
KNOWN_ISSUES
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
= Known Issues
Occasionally odd {issues}[link:ISSUES.html] arise without a transparent or
acceptable solution. Those issues are documented here.
* Some libraries/applications may install signal handlers which conflict
with signal handlers unicorn uses. Leaving "preload_app false"
(the default) will allow unicorn to always override existing signal
handlers.
* Issues with FreeBSD jails can be worked around as documented by Tatsuya Ono:
https://yhbt.net/unicorn-public/CAHBuKRj09FdxAgzsefJWotexw-7JYZGJMtgUp_dhjPz9VbKD6Q@mail.gmail.com/
* PRNGs (pseudo-random number generators) loaded before forking
(e.g. "preload_app true") may need to have their internal state
reset in the after_fork hook. Starting with unicorn 3.6.1, we
have builtin workarounds for Kernel#rand and OpenSSL::Random users,
but applications may use other PRNGs.
* For notes on sandboxing tools such as Bundler or Isolate,
see the {Sandbox}[link:Sandbox.html] page.
* nginx with "sendfile on" under FreeBSD 8 is broken when
uploads are buffered to disk. Disabling sendfile is required to
work around this bug which should be fixed in newer versions of FreeBSD.
* When using "preload_app true", with apps using background threads
need to restart them in the after_fork hook because threads are never
shared with child processes. Additionally, any synchronization
primitives (Mutexes, Monitors, ConditionVariables) should be
reinitialized in case they are held during fork time to avoid
deadlocks. The core Ruby Logger class needlessly uses a MonitorMutex
which can be disabled with a {monkey patch}[link:examples/logger_mp_safe.rb]
== Known Issues (Old)
* Under some versions of Ruby 1.8, it is necessary to call +srand+ in an
after_fork hook to get correct random number generation. We have a builtin
workaround for this starting with unicorn 3.6.1
See http://blade.nagaokaut.ac.jp/cgi-bin/scat.rb/ruby/ruby-core/36450
* On Ruby 1.8 prior to Ruby 1.8.7-p248, *BSD platforms have a broken
stdio that causes failure for file uploads larger than 112K. Upgrade
your version of Ruby or continue using unicorn 1.x/3.4.x.
* Under Ruby 1.9.1, methods like Array#shuffle and Array#sample will
segfault if called after forking. Upgrade to Ruby 1.9.2 or call
"Kernel.rand" in your after_fork hook to reinitialize the random
number generator.
See http://blade.nagaokaut.ac.jp/cgi-bin/scat.rb/ruby/ruby-core/28655
* Rails 2.3.2 bundles its own version of Rack. This may cause subtle
bugs when simultaneously loaded with the system-wide Rack Rubygem
which unicorn depends on. Upgrading to Rails 2.3.4 (or later) is
strongly recommended for all Rails 2.3.x users for this (and security
reasons). Rails 2.2.x series (or before) did not bundle Rack and are
should be unnaffected. If there is any reason which forces your
application to use Rails 2.3.2 and you have no other choice, then
you may edit your unicorn gemspec and remove the Rack dependency.
ref: https://yhbt.net/unicorn-public/20091014221552.GA30624@dcvr.yhbt.net/
Note: the workaround described in the article above only made
the issue more subtle and we didn't notice them immediately.
* WONTFIX: code reloading and restarts with Sinatra 0.3.x (and likely older
versions) apps is broken. The workaround is to force production
mode to disable code reloading as well as disabling "run" in your
Sinatra application:
set :env, :production
set :run, false
Since this is no longer an issue with Sinatra 0.9.x apps, this will not be
fixed on our end. Since unicorn is itself the application launcher, the
at_exit handler used in old Sinatra always caused Mongrel to be launched
whenever a unicorn worker was about to exit.
Also remember we're capable of replacing the running binary without dropping
any connections regardless of framework :)