Post-run error

[knightsg]

Background: I previously had an issue document in #319 where ansible using mitogen would fail immediately after starting a playbook run. That was resolved, and it appeared to be working fine. When I successfully tested that fix, I was using a playbook that imports 4 roles of 4 - 15 tasks each.

Current issue: I tried to run ansible using mitogen against our site.yml playbook that imports a bunch (18) of other playbooks. When I run site.yml against the same server (using -l ), I get this error after the run completes:

ERROR! [pid 4766] 13:57:01.410645 E mitogen.ctx.local.4790: mitogen: <IoLogger stderr> crashed
Traceback (most recent call last):
  File "<stdin>", line 1778, in _call
  File "<stdin>", line 1556, in on_receive
  File "<stdin>", line 896, in read
  File "<stdin>", line 287, in io_op
OSError: [Errno 11] Resource temporarily unavailable
ERROR! [pid 4766] 13:57:01.445036 E mitogen.ctx.local.4790: mitogen: <IoLogger stdout> crashed
Traceback (most recent call last):
  File "<stdin>", line 1778, in _call
  File "<stdin>", line 1556, in on_receive
  File "<stdin>", line 896, in read
  File "<stdin>", line 287, in io_op
OSError: [Errno 11] Resource temporarily unavailable

It doesn't appear to cause any issues apart from the error, but I thought I should post it here nonetheless.

Setup
Controller
Uname: Linux XXXXXXXXXX 4.4.0-43-Microsoft #1-Microsoft Wed Dec 31 14:42:53 PST 2014 x86_64 x86_64 x86_64 GNU/Linux (Windows Subsystem for Link - WSL, which is currently running Ubuntu 16.04.5)

Ansible:
ansible 2.4.6.0 (stable-2.4 9d08bc8242) last updated 2018/07/16 11:49:56 (GMT -700)
config file = /root/.ansible.cfg
configured module search path = [u'/root/.ansible/plugins/modules', u'/usr/share/ansible/plugins/modules']
ansible python module location = /usr/local/ansible/lib/ansible
executable location = /usr/local/ansible/bin/ansible
python version = 2.7.12 (default, Dec 4 2017, 14:50:18) [GCC 5.4.0 20160609]

Notes: Added 'strategy_plugins = /path/to/mitogen-0.2.1/ansible_mitogen/plugins/strategy' to defaults and ran ansible-playbook with 'ANSIBLE_STRATEGY=mitogen_linear' set.

Host(s)
Uname: Linux XXXXXXXXXXXXXXX 3.13.0-153-generic #203-Ubuntu SMP Thu Jun 14 08:52:28 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux
Python 2.7.6

[rouge8]

I think this is fixed in 0.2.2. http://mitogen.readthedocs.io/en/stable/changelog.html#v0-2-2-2018-07-26

[knightsg]

Unfortunately it's still happening for me in the master branch (latest pulled commit is #898c06f1).

[dw]

Sorry for the delay -- been super busy :) I'm betting this is basically an epoll manifestation of https://github.com/dw/mitogen/issues/320

The fix will be similar, however the epoll interface is totally different to kqueue, so I need some free time to properly investigate it.

There is an underlying root cause to both these bugs, the previous fix didn't find/fix it.

[knightsg]

No worries, and thanks for the reply. I'm not in any major rush to get this working.

[dw]

Somehow made it 3 months without noticing you're running WSL. The root cause is that WSL produces conflicting socket readiness / error codes when a socket is half-shutdown.

Ultra-simple fix is to avoid shutting down sockets like that on WSL, but the mechanism is required to ensure timely cleanup of child contexts.