Now it's possible to find both packages and modules when the
sys.modules[...] state for the package/module is junk. Previously only
modules were possible.
This also refactors things to make writing better tests for all these
cases much simpler.
There has always been a race in PushFileService since given a parent
asked to forward modules to two children via some intermediary:
interm = router.local()
c1 = router.local(via=interm)
c2 = router.local(via=interm)
service.propagate_to(c1, 'foo/bar.py')
service.propagate_to(c2, 'foo/bar.py')
Two calls will be emitted to 'interm':
PushFileService.store_and_forward(c1, 'foo/bar.py', [blob])
PushFileService.store(c2, 'foo/bar.py')
Which will be processed in-order up to the point where service pool
threads in 'interm' are woken to process the message.
While it is guaranteed store_and_forward() will be processed first, no
guarantee existed that its assigned pool thread would wake and take
_lock first, thus it was possible for forward() to win the race, and for
a request to arrive to forward a file that had not been placed in local
cache yet.
Here we get rid of SerializedInvoker entirely, as it is partially to
blame for hiding the race: SerializedInvoker can only ensure no two
messages are processed simultaneously, it cannot ensure the messages are
processed in their intended order.
Instead, teach forward() that it may be called before
store_and_forward(), and if that is the case, to place the forward
request on to _waiters alongside any local threads blocked in get().
Regardless of the version of simplejson loaded in the master, load up
the ModuleResponder cache with our 2.4-compatible version.
To cope with simplejson being loaded due to modules like ec2_group that
try to import it before importing 'json', also update target.py to
remove it from the whitelist if a local 'json' module import succeeds.
Since 802de6a8d5, sudo on CentOS 5 had
begun failing due to a TTY FD leak in the parent process being fixed.
The old versions of sudo doesn't hang around after starting a child --
they exec the privilege-escalated child process on top of themselves,
meaning no spare copy of the TTY FD is kept alive by sudo.
When the child starts up, it replaces stdio with IoLoggers, including
the inherited stderr FD connected to DiagLogStream/the slave PTY. When
the last process closes a slave PTY, the kernel sends SIGHUP to any
processes still having it as the controlling TTY.
Therefore we must either ignore SIGHUP until the first stage has been
waited on (since the first stage also preserve the FD), or dup the
inherited TTY FD and keep it around forever.
Wasting one FD seems less annoying than modifying process signals for
all potential library users, so that is the approach taken here.
[costapp]
ERROR! [pid 25135] 21:10:56.284733 E mitogen.ctx.ssh.35.200.203.48: mitogen: While calling no-reply method PushFileService.forward
Traceback (most recent call last):
File "master:/home/dmw/src/mitogen/mitogen/service.py", line 260, in _invoke
ret = method(**kwargs)
File "master:/home/dmw/src/mitogen/mitogen/service.py", line 718, in forward
self._forward(path, context)
File "master:/home/dmw/src/mitogen/mitogen/service.py", line 633, in _forward
stream = self.router.stream_by_id(context.context_id)
AttributeError: 'unicode' object has no attribute 'context_id'
^C [ERROR]: User interrupted execution
Minify-safe files are marked with a magical "# !mitogen: minify_safe"
comment anywhere in the file, which activates the minifier. The result
is naturally cached by ModuleResponder, therefore lru_cache is gone too.
Given:
import os, mitogen
@mitogen.main()
def main(router):
c = router.ssh(hostname='k3')
c.call(os.getpid)
router.sudo(via=c)
SSH footprint drops from 56.2 KiB to 42.75 KiB (-23.9%)
Ansible "shell: hostname" drops 149.26 KiB to 117.42 KiB (-21.3%)
When the interpreter is modern enough, use zlib.compressobj() to
pre-compress the unchanging parts of the bootstrap once, then use
compressobj.copy() to append just the context's config during stream
construction.
Before: 100 loops, best of 3: 5.81 msec per loop
After: 10000 loops, best of 3: 35.9 usec per loop
With 100 targets this is enough to knock 6 seconds off startup, at 500
targets it becomes half a minute.
Test 'program':
python -m timeit -s '
import mitogen.parent as p;
import mitogen.master as m;
r=m.Router();
s=p.Stream(r, 0, max_message_size=1);
r.broker.shutdown()'\
\
's.get_preamble()'
Ansible modules were being resent continuously - but only the main
script module, and any custom modutils if any were present.
Wire footprint drops by ~1/3rd for a 500 task run of 'shell: hostname':
-rw-r--r-- 1 root root 584K Jan 31 22:06 500mito-before2
-rw-r--r-- 1 root root 434K Jan 31 22:04 500mito-filesbugonly
Single task 100 SSH target run, before:
3533181 function calls (3533083 primitive calls) in 616.688 seconds
User time (seconds): 32.52
System time (seconds): 2.71
Percent of CPU this job got: 64%
Elapsed (wall clock) time (h:mm:ss or m:ss): 0:54.88
After:
451602 function calls (451504 primitive calls) in 570.746 seconds
User time (seconds): 29.48
System time (seconds): 2.81
Percent of CPU this job got: 67%
Elapsed (wall clock) time (h:mm:ss or m:ss): 0:48.20
David Wilson