Date | Commit message (Collapse) |
|
This release contains fairly major internal workings of master-to-worker
notifications. The master process no longer sends signals to workers
for most tasks. This works around some compatibility issues with some
versions of the "pg" gem (and potentially any other code which may not
handle EINTR properly). One extra benefit is it also helps stray
workers notice a rare, unexpected master death more easily. Workers
continue to (and will always) accept existing signals for compatibility
with tools/scripts which may signal workers.
PID file are always written early (even on upgrade) again to avoid
breaking strange monitoring setups which use PID files. Keep in mind we
have always discouraged monitoring based on PID files as they are
fragile.
We now avoid bubbling IOError to the Rack app on premature client
disconnects when streaming the input body. This is usually not a
problem with nginx, but may be on some LAN setups without nginx).
Thanks to Sam Saffron, Jimmy Soho, Rodrigo Rosenfeld Rosas,
Michael Fischer, and Andrew Hobson for their help with this release.
Note: the unicorn mailing list will be moved/changed soon due to the
RubyForge shutdown. unicorn will always rely only on Free Software.
There will never be any sign-up requirements nor terms-of-service to
agree to when communicating with us.
|
|
"check" rolls off the fingers of users familiar with GNU Automake
more easily. The "test-all" target is preserved for compatibility.
|
|
Eric Wong (6):
tests: fix SO_REUSEPORT tests for old Linux and non-Linux
stream_input: avoid IO#close on client disconnect
t0300: kill off stray processes in test
always write PID file early for compatibility
doc: clarify SIGNALS and reference init example
rework master-to-worker signaling to use a pipe
|
|
Signaling using normal kill(2) is preserved, but the master now
prefers to signal workers using a pipe rather than kill(2).
Non-graceful signals (:TERM/:KILL) are still sent using kill(2),
as they ask for immediate shutdown.
This change is necessary to avoid triggering the ubf (unblocking
function) for rb_thread_call_without_gvl (and similar) functions
extensions. Most notably, this fixes compatibility with newer
versions of the 'pg' gem which will cancel a running DB query if
signaled[1].
This also has the nice side-effect of allowing a premature
master death (assuming preload_app didn't cause the master to
spawn off rogue child daemons).
Note: users should also refrain from using "killall" if using the
'pg' gem or something like it.
Unfortunately, this increases FD usage in the master as the writable
end of the pipe is preserved in the master. This limit the number
of worker processes the master may run to the open file limit of the
master process. Increasing the open file limit of the master
process may be needed. However, the FD use on the workers is
reduced by one as the internal self-pipe is no longer used. Thus,
overall pipe allocation for the kernel remains unchanged.
[1] - pg is correct to cancel a query, as it cannot know if
the signal was for a) graceful unicorn shutdown or
b) oh-noes-I-started-a-bad-query-ABORT-ABORT-ABORT!!
|
|
"interactive terminal" needed clarification.
While we're at it, link to the init.sh example since it may
be shared with nginx.
Reported-by: Rodrigo Rosenfeld Rosas
ref: <5294E9D4.5030608@gmail.com>
|
|
This reduces the window for a non-existent PID for folks who monitor
PIDs (not a great idea anyways). Unfortunately, this change also brings
us back to the case where having a PID later (for other process monitors)
is beneficial but more unicorn releases exist where we write the PID
early.
Thanks to Jimmy Soho for reporting this issue.
ref: <CAHStS5gFYcPBDxkVizAHrOeDKAkjT69kruFdgaY0CbB+vLbK8Q@mail.gmail.com>
This partially reverts 7d6ac0c17eb29a00a5b74099dbb3d4d015999f27
Folks: please monitor your app with HTTP requests rather than checking
processes, a stuck/wedged Ruby VM is still a running one.
|
|
We shouldn't leave processes running after the test.
|
|
This can avoid IOError from being seen by the application, and also
reduces points where IO#close may be called. This is a good thing
if we eventually port this code into a low-level server like
cmogstored where per-client memory space is defined by FD number of
a client.
Reported-by: Andrew Hobson <ahobson@gmail.com>
|
|
On BSD-derived platforms the getsockopt true value may be any
(>= 0) value, not just one as it is on Linux.
Additionally, SO_REUSEPORT is only supported since Linux 3.9, so
folks on older kernels may not have it available. We still define it
for Linux since kernel upgrades are usually more common than glibc
upgrades.
Note: we will still raise an exception at runtime if a user
explicitly requests :reuseport in their config and runs an
older Linux kernel.
Reported-by: Andrew Hobson <ahobson@gmail.com>
|
|
* support SO_REUSEPORT on new listeners (:reuseport)
This allows users to start an independent instance of unicorn on
a the same port as a running unicorn (as long as both instances
use :reuseport).
ref: https://lwn.net/Articles/542629/
* unicorn is now GPLv2-or-later and Ruby 1.8-licensed
(instead of GPLv2-only, GPLv3-only, and Ruby 1.8-licensed)
This changes nothing at the moment. Once the FSF publishes the next
version of the GPL, users may choose the newer GPL version without the
unicorn BDFL approving it. Two years ago when I got permission to add
GPLv3 to the license options, I also got permission from all past
contributors to approve future versions of the GPL. So now I'm
approving all future versions of the GPL for use with unicorn.
Reasoning below:
In case the GPLv4 arrives and I am not alive to approve/review it,
the lesser of evils is have give blanket approval of all future GPL
versions (as published by the FSF). The worse evil is to be stuck
with a license which cannot guarantee the Free-ness of this project
in the future.
This unfortunately means the FSF can theoretically come out with
license terms I do not agree with, but the GPLv2 and GPLv3 will
always be an option to all users.
Note: we currently prefer GPLv3
Two improvements thanks to Ernest W. Durbin III:
* USR2 redirects fixed for Ruby 1.8.6 (broken since 4.1.0)
* unicorn(1) and unicorn_rails(1) enforces valid integer for -p/--port
A few more odd, minor tweaks and fixes:
* attempt to rename PID file when possible (on USR2)
* workaround reopen atomicity issues for stdio vs non-stdio
* improve handling of client-triggerable socket errors
|
|
Users may confuse '-p' with the (to-be-deprecated) '-P/--pid'
option, leading to surprising behavior if a pathname is passed as a
port, because String#to_i would convert it to zero, causing:
TCPServer.new(host, port = 0)
to bind to a random, unused port.
|
|
This renables the ability for Ruby 1.8.6 environments to perform reexecs
[ew: clarified this is for 1.8.6,
favor literal {} over Hash.new,
tweaked LISTENERS.map => LISTENERS.each, thanks to Hleb Valoshka
]
Signed-off-by: Eric Wong <normalperson@yhbt.net>
|
|
In case we (and Linux) supports other values in the future,
we can update it then. Until now, ensure users only set
true or false for this option.
|
|
There is currently no GPLv4, so this change has no effect at the
moment.
In case the GPLv4 arrives and I am not alive to approve/review it,
the lesser of evils is have give blanket approval of all future GPL
versions (as published by the FSF). The worse evil is to be stuck
with a license which cannot guarantee the Free-ness of this project
in the future.
This unfortunately means the FSF can theoretically come out with
license terms I do not agree with, but the GPLv2 and GPLv3 will
always be an option to all users.
|
|
Thanks to Hongli Lai for noticing my typo. While we're at it, finish up
a halfway-written comment for the EXDEV case
|
|
Ruby 1.9 and later includes IO#autoclose=, so we can use it
and prevent some dead IO objects from hanging around.
|
|
This allows users to start an independent instance of unicorn on
a the same port as a running unicorn (as long as both instances
use :reuseport).
ref: https://lwn.net/Articles/542629/
|
|
Otherwise these tests fail if we start using IO#autoclose=true
on Ruby 1.9 (and also if we use IPv6 sockets for tests).
|
|
This will preserve mtime on successful renames for comparisions.
While we're at it, avoid writing the new PID until the listeners are
inherited successfully. This can be useful to avoid accidentally
clobbering a good PID if binding the listener or building the app
(preload_app==true) fails
|
|
In multithreaded apps, we must use dup2/dup3 with a temporary
descriptor to reopen log files atomically. This is the only way
to protect all concurrent userspace access to a file when reopening.
ref: http://bugs.ruby-lang.org/issues/9036
ref: yahns commit bcb10abe53cfb1d6a8ef7daef59eb10ced397c8a
|
|
RAA is dead.
|
|
All tests seem to pass.
|
|
Thanks to Eric Chapweske for the heads up.
ref: http://mid.gmane.org/loom.20130904T205308-432@post.gmane.org
|
|
As of r40610 in ruby trunk, internal encoding is ignored if
external coding is ASCII-8BIT (binary)
ref: r40610 http://svn.ruby-lang.org/repos/ruby/trunk
|
|
We do not attempt to write HTTP responses for socket errors if
clients disconnect from us unexpectedly.
Additionally, we do not hide backtraces EINVAL/EBADF errors, since
they are indicative of real bugs which must be fixed.
We do continue to hide hide EOF, ECONNRESET, ENOTCONN, and EPIPE
because clients (even "friendly") ones will break connections due to
client crashes or network failure (which is common for me :P), and
the backtraces from those will cause excessive logging and even
become a DoS vector.
|
|
Thanks to Micah Chalmer for this fix. There are also minor
documentation updates and internal cleanups.
|
|
This fixes the -N (a.k.a. --no-defaut-middleware) option, which
was not working. The problem was that Unicorn::Configurator::RACKUP
is cleared before the lambda returned by Unicorn.builder is run,
which means that checking whether the :no_default_middleware option
was set from the lambda could not detect anything. This patch copies
it to a local variable that won't get clobbered, restoring the feature.
[ew: squashed test commit into the fix, whitespace fixes]
Signed-off-by: Eric Wong <normalperson@yhbt.net>
|
|
This could allow servers with persistent connection support[1]
to support our check_client_connection in the future.
[1] - Rainbows!/zbatery, possibly others
|
|
Otherwise, the signalled process may take too long to react to
and process all the signals on machines with few CPUs.
|
|
The PrivateTmp feature of systemd breaks the usage of /tmp for the
shared Unix domain socket between nginx and unicorn, so discourage the
use of /tmp in that case.
While we're at it, use consistent paths for everything and use an
obviously intended-for-user-customization "/path/to" prefix instead
of "/tmp"
ML-Ref: CAKLVLx_t+9zWMhquMWDfStrxS7xrNoGmN0ZDsjSCUE=VxU+oyQ@mail.gmail.com
Reported-by: David Wilkins <dwilkins@conecuh.com>
|
|
This release fixes a bug in Unicorn::HttpParser#filter_body
which affected some configurations of Rainbows! There is
also a minor size reduction in the DSO.
|
|
Our rb_str_modify() became no-ops due to incomplete reverts
of workarounds for old Rubinius, causing rb_str_set_len to
fail with: can't set length of shared string (RuntimeError)
This bug was introduced due to improper workarounds for old
versions of Rubinius in 2009 and 2010:
commit 5e8979ad38efdc4de3a69cc53aea33710d478406
("http: cleanups for latest Rubinius")
commit f37c23704cb73d57e9e478295d1641df1d9104c7
("http: no-op rb_str_modify() for Rubies without it")
|
|
Extra pointers waste space in the DSO. Normally I wouldn't
care, but the string lengths are identical and this code
already made it into another project in this form.
size(1) output:
text data bss dec hex filename
before: 42881 2040 336 45257 b0c9 unicorn_http.so
after: 42499 1888 336 44723 aeb3 unicorn_http.so
ref: http://www.akkadia.org/drepper/dsohowto.pdf
|
|
Unicorn::Const::UNICORN_VERSION is now auto-generated from
GIT-VERSION-GEN and always correct. Minor cleanups for
hijacking.
|
|
As far as I can tell, this was never necessary.
|
|
commit a9474624a148fe58e0944664190b259787dcf51e in rack.git
|
|
This DRYs out our code and prevents snafus like the 4.6.0
release where UNICORN_VERSION stayed at 4.5.0
Reported-by: Maurizio De Santis <m.desantis@morganspa.com>
|
|
This pre-release adds hijacking support for Rack 1.5 users.
See Rack documentation for more information about hijacking.
There is also a new --no-default-middleware/-N option
for the `unicorn' command to ignore RACK_ENV within unicorn
thanks to Lin Jen-Shin.
There are only documentation and test-portability updates
since 4.6.0pre1, no code changes.
|
|
On FreeBSD 9.0, "wc -l" emits leading whitespace, so
filter it through tr -d '[:space:]' to eliminate it.
|
|
On FreeBSD 9.0, "wc -c" emits leading whitespace, so
filter it through tr -d '[:space:]' to eliminate it.
This is commit 8a6117a22a7d01eeb5adc63d3152acf435cd3176
in rainbows.git
|
|
"date +%s" is not in POSIX (it is in GNU, and at least FreeBSD
9.0, possibly earlier). The Ruby equivalent should be
sufficiently portable between different Ruby versions.
This change was automated via:
perl -i -p -e 's/date \+%s/unix_time/' t/*.sh
This is commit 0ba6fc3c30b9cf530faf7fcf5ce7be519ec13fe7
in rainbows.git
|
|
POSIX already stipulates tee(1) must be unbuffered. I think my
decision to use utee was due to my being misled by a bug in
older curl where -N did not work as advertised (but --no-buffer
did).
N.B. we don't use tee in unicorn tests, this just matches
commit cbff7b0892148b037581541184364e0e91d2a138 in rainbows
|
|
-N/--no-default-middleware needs a corresponding manpage entry.
Additionally, the Rack::Chunked/ContentLength middleware comment
is out-of-date as of unicorn v4.1.0
|
|
This pre-release adds hijacking support for Rack 1.5 users.
See Rack documentation for more information about hijacking.
There is also a new --no-default-middleware/-N option
for the `unicorn' command to ignore RACK_ENV within unicorn.
|
|
* hijack:
ignore normal Rack response at request-time hijack
support for Rack hijack in request and response
|
|
This would prevent Unicorn from adding default middleware,
as if RACK_ENV were always none. (not development nor deployment)
This should also be applied to `rainbows' and `zbatery' as well.
One of the reasons to add this is to avoid conflicting
RAILS_ENV and RACK_ENV. It would be helpful in the case
where a Rails application and Rack application are composed
together, while we want Rails app runs under development
and Rack app runs under none (if we don't want those default
middleware), and we don't really want to make RAILS_ENV
set to development and RACK_ENV to none because it might be
confusing. Note that Rails would also look into RACK_ENV.
Another reason for this is that only `rackup' would be
inserting those default middleware. Both `thin' and `puma'
would not do this, nor does Rack::Handler.get.run which is
used in Sinatra.
So using this option would make it work differently from
`rackup' but somehow more similar to `thin' or `puma'.
Discussion thread on the mailing list:
http://rubyforge.org/pipermail/mongrel-unicorn/2013-January/001675.html
Signed-off-by: Eric Wong <normalperson@yhbt.net>
|
|
This off-by-one error was incorrectly rejecting a line which
would've been readable without wrapping on an 80-column terminal.
|
|
This fixes a Rack::Lint regression discovered in t0005.
|
|
Once a connection is hijacked, we ignore it completely and leave
the connection at the mercy of the application.
|
|
Rack 1.5.0 (protocol version [1,2]) adds support for
hijacking the client socket (removing it from the control
of unicorn (or any other Rack webserver)).
Tested with rack 1.5.0.
|