##// END OF EJS Templates
wireproto: add streams to frame-based protocol...
wireproto: add streams to frame-based protocol Previously, the frame-based protocol was just a series of frames, with each frame associated with a request ID. In order to scale the protocol, we'll want to enable the use of compression. While it is possible to enable compression at the socket/pipe level, this has its disadvantages. The big one is it undermines the point of frames being standalone, atomic units that can be read and written: if you add compression above the framing protocol, you are back to having a stream-based protocol as opposed to something frame-based. So in order to preserve frames, compression needs to occur at the frame payload level. Compressing each frame's payload individually will limit compression ratios because the window size of the compressor will be limited by the max frame size, which is 32-64kb as currently defined. It will also add CPU overhead, as it is more efficient for compressors to operate on fewer, larger blocks of data than more, smaller blocks. So compressing each frame independently is out. This means we need to compress each frame's payload as if it is part of a larger stream. The simplest approach is to have 1 stream per connection. This could certainly work. However, it has disadvantages (documented below). We could also have 1 stream per RPC/command invocation. (This is the model HTTP/2 goes with.) This also has disadvantages. The main disadvantage to one global stream is that it has the very real potential to create CPU bottlenecks doing compression. Networks are only getting faster and the performance of single CPU cores has been relatively flat. Newer compression formats like zstandard offer better CPU cycle efficiency than predecessors like zlib. But it still all too common to saturate your CPU with compression overhead long before you saturate the network pipe. The main disadvantage with streams per request is that you can't reap the benefits of the compression context for multiple requests. For example, if you send 1000 RPC requests (or HTTP/2 requests for that matter), the response to each would have its own compression context. The overall size of the raw responses would be larger because compression contexts wouldn't be able to reference data from another request or response. The approach for streams as implemented in this commit is to support N streams per connection and for streams to potentially span requests and responses. As explained by the added internals docs, this facilitates servers and clients delegating independent streams and compression to independent threads / CPU cores. This helps alleviate the CPU bottleneck of compression. This design also allows compression contexts to be reused across requests/responses. This can result in improved compression ratios and less overhead for compressors and decompressors having to build new contexts. Another feature that was defined was the ability for individual frames within a stream to declare whether that individual frame's payload uses the content encoding (read: compression) defined by the stream. The idea here is that some servers may serve data from a combination of caches and dynamic resolution. Data coming from caches may be pre-compressed. We want to facilitate servers being able to essentially stream bytes from caches to the wire with minimal overhead. Being able to mix and match with frames are compressed within a stream enables these types of advanced server functionality. This commit defines the new streams mechanism. Basic code for supporting streams in frames has been added. But that code is seriously lacking and doesn't fully conform to the defined protocol. For example, we don't close any streams. And support for content encoding within streams is not yet implemented. The change was rather invasive and I didn't think it would be reasonable to implement the entire feature in a single commit. For the record, I would have loved to reuse an existing multiplexing protocol to build the new wire protocol on top of. However, I couldn't find a protocol that offers the performance and scaling characteristics that I desired. Namely, it should support multiple compression contexts to facilitate scaling out to multiple CPU cores and compression contexts should be able to live longer than single RPC requests. HTTP/2 *almost* fits the bill. But the semantics of HTTP message exchange state that streams can only live for a single request-response. We /could/ tunnel on top of HTTP/2 streams and frames with HEADER and DATA frames. But there's no guarantee that HTTP/2 libraries and proxies would allow us to use HTTP/2 streams and frames without the HTTP message exchange semantics defined in RFC 7540 Section 8. Other RPC protocols like gRPC tunnel are built on top of HTTP/2 and thus preserve its semantics of stream per RPC invocation. Even QUIC does this. We could attempt to invent a higher-level stream that spans HTTP/2 streams. But this would be violating HTTP/2 because there is no guarantee that HTTP/2 streams are routed to the same server. The best we can do - which is what this protocol does - is shoehorn all request and response data into a single HTTP message and create streams within. At that point, we've defined a Content-Type in HTTP parlance. It just so happens our media type can also work as a standalone, stream-based protocol, without leaning on HTTP or similar protocol. Differential Revision: https://phab.mercurial-scm.org/D2907

File last commit:

r37233:d2bd29df default
r37304:9bfcbe4f default
Show More
server.py
211 lines | 6.8 KiB | text/x-python | PythonLexer
Yuya Nishihara
server: move cmdutil.service() to new module (API)...
r30506 # server.py - utility and factory of server
#
# Copyright 2005-2007 Matt Mackall <mpm@selenic.com>
#
# This software may be used and distributed according to the terms of the
# GNU General Public License version 2 or any later version.
from __future__ import absolute_import
import os
import tempfile
from .i18n import _
from . import (
Yuya Nishihara
chgserver: make it a core module and drop extension flags...
r30513 chgserver,
Matt Harbison
serve: add support for Mercurial subrepositories...
r32005 cmdutil,
Yuya Nishihara
server: move service table and factory from commandserver...
r30507 commandserver,
Yuya Nishihara
server: move cmdutil.service() to new module (API)...
r30506 error,
Yuya Nishihara
server: move service factory from hgweb
r30509 hgweb,
Augie Fackler
server: use pycompat to get argv
r32530 pycompat,
Yuya Nishihara
server: move cmdutil.service() to new module (API)...
r30506 util,
)
Yuya Nishihara
procutil: bulk-replace util.std* to point to new module
r37137 from .utils import (
procutil,
)
Yuya Nishihara
server: move cmdutil.service() to new module (API)...
r30506 def runservice(opts, parentfn=None, initfn=None, runfn=None, logfile=None,
runargs=None, appendpid=False):
'''Run a command as a service.'''
Matt Harbison
server: refactor 'daemon_postexec' instructions into a dictionary
r37232 postexecargs = {}
if opts['daemon_postexec']:
for inst in opts['daemon_postexec']:
if inst.startswith('unlink:'):
postexecargs['unlink'] = inst[7:]
elif inst.startswith('chdir:'):
postexecargs['chdir'] = inst[6:]
elif inst != 'none':
raise error.Abort(_('invalid value for --daemon-postexec: %s')
% inst)
Matt Harbison
server: add an error feedback mechanism for when the daemon fails to launch...
r37229 # When daemonized on Windows, redirect stdout/stderr to the lockfile (which
# gets cleaned up after the child is up and running), so that the parent can
# read and print the error if this child dies early. See 594dd384803c. On
# other platforms, the child can write to the parent's stdio directly, until
# it is redirected prior to runfn().
if pycompat.iswindows and opts['daemon_postexec']:
Matt Harbison
server: refactor 'daemon_postexec' instructions into a dictionary
r37232 if 'unlink' in postexecargs and os.path.exists(postexecargs['unlink']):
procutil.stdout.flush()
procutil.stderr.flush()
Matt Harbison
server: add an error feedback mechanism for when the daemon fails to launch...
r37229
Matt Harbison
server: refactor 'daemon_postexec' instructions into a dictionary
r37232 fd = os.open(postexecargs['unlink'],
os.O_WRONLY | os.O_APPEND | os.O_BINARY)
try:
Matt Harbison
server: minor code cleanup...
r37233 os.dup2(fd, procutil.stdout.fileno())
os.dup2(fd, procutil.stderr.fileno())
Matt Harbison
server: refactor 'daemon_postexec' instructions into a dictionary
r37232 finally:
os.close(fd)
Matt Harbison
server: add an error feedback mechanism for when the daemon fails to launch...
r37229
Yuya Nishihara
server: move cmdutil.service() to new module (API)...
r30506 def writepid(pid):
if opts['pid_file']:
if appendpid:
Augie Fackler
server: write out pid using bytes IO instead of str IO
r32548 mode = 'ab'
Yuya Nishihara
server: move cmdutil.service() to new module (API)...
r30506 else:
Augie Fackler
server: write out pid using bytes IO instead of str IO
r32548 mode = 'wb'
Yuya Nishihara
server: move cmdutil.service() to new module (API)...
r30506 fp = open(opts['pid_file'], mode)
Yuya Nishihara
py3: simply use b'%d\n' to format pid in server.py...
r32617 fp.write('%d\n' % pid)
Yuya Nishihara
server: move cmdutil.service() to new module (API)...
r30506 fp.close()
if opts['daemon'] and not opts['daemon_postexec']:
# Signal child process startup with file removal
lockfd, lockpath = tempfile.mkstemp(prefix='hg-service-')
os.close(lockfd)
try:
if not runargs:
Yuya Nishihara
procutil: bulk-replace function calls to point to new module
r37138 runargs = procutil.hgcmd() + pycompat.sysargv[1:]
Yuya Nishihara
server: move cmdutil.service() to new module (API)...
r30506 runargs.append('--daemon-postexec=unlink:%s' % lockpath)
# Don't pass --cwd to the child process, because we've already
# changed directory.
for i in xrange(1, len(runargs)):
if runargs[i].startswith('--cwd='):
del runargs[i]
break
elif runargs[i].startswith('--cwd'):
del runargs[i:i + 2]
break
def condfn():
return not os.path.exists(lockpath)
Yuya Nishihara
procutil: bulk-replace function calls to point to new module
r37138 pid = procutil.rundetached(runargs, condfn)
Yuya Nishihara
server: move cmdutil.service() to new module (API)...
r30506 if pid < 0:
Matt Harbison
server: add an error feedback mechanism for when the daemon fails to launch...
r37229 # If the daemonized process managed to write out an error msg,
# report it.
if pycompat.iswindows and os.path.exists(lockpath):
Matt Harbison
server: minor code cleanup...
r37233 with open(lockpath, 'rb') as log:
Matt Harbison
server: add an error feedback mechanism for when the daemon fails to launch...
r37229 for line in log:
procutil.stderr.write(line)
Yuya Nishihara
server: move cmdutil.service() to new module (API)...
r30506 raise error.Abort(_('child process failed to start'))
writepid(pid)
finally:
Ryan McElroy
server: use tryunlink
r31548 util.tryunlink(lockpath)
Yuya Nishihara
server: move cmdutil.service() to new module (API)...
r30506 if parentfn:
return parentfn(pid)
else:
return
if initfn:
initfn()
if not opts['daemon']:
Yuya Nishihara
procutil: bulk-replace function calls to point to new module
r37138 writepid(procutil.getpid())
Yuya Nishihara
server: move cmdutil.service() to new module (API)...
r30506
if opts['daemon_postexec']:
try:
os.setsid()
except AttributeError:
pass
Matt Harbison
server: add an error feedback mechanism for when the daemon fails to launch...
r37229
Matt Harbison
server: refactor 'daemon_postexec' instructions into a dictionary
r37232 if 'chdir' in postexecargs:
os.chdir(postexecargs['chdir'])
Yuya Nishihara
procutil: bulk-replace function calls to point to new module
r37138 procutil.hidewindow()
Yuya Nishihara
procutil: bulk-replace util.std* to point to new module
r37137 procutil.stdout.flush()
procutil.stderr.flush()
Yuya Nishihara
server: move cmdutil.service() to new module (API)...
r30506
nullfd = os.open(os.devnull, os.O_RDWR)
logfilefd = nullfd
if logfile:
Yuya Nishihara
server: drop executable bit from daemon log file...
r34925 logfilefd = os.open(logfile, os.O_RDWR | os.O_CREAT | os.O_APPEND,
0o666)
Matt Harbison
server: minor code cleanup...
r37233 os.dup2(nullfd, procutil.stdin.fileno())
os.dup2(logfilefd, procutil.stdout.fileno())
os.dup2(logfilefd, procutil.stderr.fileno())
stdio = (procutil.stdin.fileno(), procutil.stdout.fileno(),
procutil.stderr.fileno())
if nullfd not in stdio:
Yuya Nishihara
server: move cmdutil.service() to new module (API)...
r30506 os.close(nullfd)
Matt Harbison
server: minor code cleanup...
r37233 if logfile and logfilefd not in stdio:
Yuya Nishihara
server: move cmdutil.service() to new module (API)...
r30506 os.close(logfilefd)
Matt Harbison
server: add an error feedback mechanism for when the daemon fails to launch...
r37229 # Only unlink after redirecting stdout/stderr, so Windows doesn't
# complain about a sharing violation.
Matt Harbison
server: refactor 'daemon_postexec' instructions into a dictionary
r37232 if 'unlink' in postexecargs:
os.unlink(postexecargs['unlink'])
Matt Harbison
server: add an error feedback mechanism for when the daemon fails to launch...
r37229
Yuya Nishihara
server: move cmdutil.service() to new module (API)...
r30506 if runfn:
return runfn()
Yuya Nishihara
server: move service table and factory from commandserver...
r30507
_cmdservicemap = {
Yuya Nishihara
chgserver: make it a core module and drop extension flags...
r30513 'chgunix': chgserver.chgunixservice,
Yuya Nishihara
server: move service table and factory from commandserver...
r30507 'pipe': commandserver.pipeservice,
'unix': commandserver.unixforkingservice,
}
Yuya Nishihara
server: add public function to select either cmdserver or hgweb
r30510 def _createcmdservice(ui, repo, opts):
Yuya Nishihara
server: move service table and factory from commandserver...
r30507 mode = opts['cmdserver']
try:
return _cmdservicemap[mode](ui, repo, opts)
except KeyError:
raise error.Abort(_('unknown mode %s') % mode)
Yuya Nishihara
server: move service factory from hgweb
r30509
Yuya Nishihara
server: add public function to select either cmdserver or hgweb
r30510 def _createhgwebservice(ui, repo, opts):
Yuya Nishihara
server: move service factory from hgweb
r30509 # this way we can check if something was given in the command-line
if opts.get('port'):
opts['port'] = util.getport(opts.get('port'))
Martin von Zweigbergk
cleanup: use set literals...
r32291 alluis = {ui}
Yuya Nishihara
server: move service factory from hgweb
r30509 if repo:
baseui = repo.baseui
alluis.update([repo.baseui, repo.ui])
else:
baseui = ui
webconf = opts.get('web_conf') or opts.get('webdir_conf')
if webconf:
Matt Harbison
serve: add support for Mercurial subrepositories...
r32005 if opts.get('subrepos'):
raise error.Abort(_('--web-conf cannot be used with --subrepos'))
Yuya Nishihara
server: move service factory from hgweb
r30509 # load server settings (e.g. web.port) to "copied" ui, which allows
# hgwebdir to reload webconf cleanly
servui = ui.copy()
servui.readconfig(webconf, sections=['web'])
alluis.add(servui)
Matt Harbison
serve: add support for Mercurial subrepositories...
r32005 elif opts.get('subrepos'):
servui = ui
# If repo is None, hgweb.createapp() already raises a proper abort
# message as long as webconf is None.
if repo:
webconf = dict()
cmdutil.addwebdirpath(repo, "", webconf)
Yuya Nishihara
server: move service factory from hgweb
r30509 else:
servui = ui
optlist = ("name templates style address port prefix ipv6"
" accesslog errorlog certificate encoding")
for o in optlist.split():
val = opts.get(o, '')
if val in (None, ''): # should check against default options instead
continue
for u in alluis:
u.setconfig("web", o, val, 'serve')
app = hgweb.createapp(baseui, repo, webconf)
return hgweb.httpservice(servui, app, opts)
Yuya Nishihara
server: add public function to select either cmdserver or hgweb
r30510
def createservice(ui, repo, opts):
if opts["cmdserver"]:
return _createcmdservice(ui, repo, opts)
else:
return _createhgwebservice(ui, repo, opts)