msg: make async backend default

[yuyuyu101]

Signed-off-by: Haomai Wang haomai@xsky.com

[yuyuyu101]

@yuriw @tchaikov and others... I hope anyone who ran a lot of qa suites can pick up this pr.

[yuyuyu101]

http://pulpito.ceph.com/haomai-2016-08-17_01:23:30-rados-wip-async-default-distro-basic-mira/

[yuriw]

@yuyuyu101 it will be part of wip-yuri-testing2_2016_08_22 tests, pls remove DNM as you see fit

[yuriw]

@yuyuyu101 can this be merged ?
Test run passed http://pulpito.ceph.com/yuriw-2016-08-26_22:32:43-rados-wip-yuri-testing_2016_08_26-distro-basic-smithi

[yuyuyu101]

@liewegas @athanatos would you think this is a good time point?

[yuriw]

Caution - we saw some unit tests failing which we (Josh and me)
initially thought were not related to the PRs (and this one in
particular), and I'm not so confident now as all tests verified passed
on master yesterday.

1 - test-ceph-helpers.sh (Failed)
5 - cephtool-test-mon.sh (Failed)
14 - run-tox-ceph-disk (Failed)
15 - run-tox-ceph-detect-init (Failed)

[liewegas]

assuming the unit test failures aren't related to async msgr (seems unlikely!) i'm all for switching the default again. iirc there was a cephfs deadlock issue.. did that get fixed?

[yuyuyu101]

yes, I don't think the uniitests failed related to async msgr too. I guess cephfs's problem is related to process spawn? @tchaikov @jcsp

[yuriw]

@jdurgin FYI ^

[jcsp]

Is there a log of the cephfs failure mentioned here?

[liewegas]

I just remember hearing something about it a few months ago. IIRC fast
dispatch related? @gregsfortytwo knows I think

[gregsfortytwo]

It wouldn't have been fast dispatch for the FS; I think it was the command with the bad locking that Doug fixed.

[yuriw]

unit tests failing b/c incorrect reg exp match issue, so tests actually passed per @jdurgin

related fix #10943

[yuriw]

I am reclaiming wip-yuri-teting tag, place it back to queue if it needs retesting retesting

[yuyuyu101]

Hmm, I guess we could merge now.