On Jan 8, 2016, at 7:29 AM, FUJITA Tomonori <[email protected]> wrote:
> 
> Had better to release 3.29.1 on pypi for OpenStack?

Would there be room for a major stability fix, or better to wait for 3.30?

I have been testing extensively at Duke; it appears that, under high load, 
eventlet queues are *not* synch-safe.
I have put a semaphore into place to protect access to the queue, as well as a 
raft of other fixes to prevent leaked file descriptors.

Background:
We have some Arista switches running an engineering build, and have seen some 
odd behavior out of them.
That led to investigations using tcpkill against port 6633 and iptables drops 
on port 6633 (to simulate an attacker interrupting the control channel with FIN 
packets, and to simulate an instantaneous control channel cut).

We *finally* have what we believe is a good fix - but if you would prefer more 
time for it to “burn-in” in our environment, I completely understand.
I had otherwise been planning to submit it around January 25th, after an 
extensive burn-in period, just slightly in advance of the 3.30 release.

Best,
Victor
--
Victor J. Orlikowski <> vjo@[cs.]duke.edu

------------------------------------------------------------------------------
_______________________________________________
Ryu-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ryu-devel

Reply via email to