On Mon, Oct 8, 2012 at 4:18 PM, Ian Barber <[email protected]> wrote:
> On Mon, Oct 8, 2012 at 8:49 PM, Thomas S Hatch <[email protected]> wrote: > > Nope, we get no errors, things just stop talking. The tricky thing too is > > that it varies widely on the infrastructure using Salt. You may notice > that > > on the pub side we have the pub_refresh stuff that refreshed the > publisher, > > this works wonders on cloud infrastructures but breaks local > > infrastructures, and on the sub side we have the sub_timeout stuff which > > help in most cases but is not needed in all of them. > > > > So we have all been running Salt with the "right mix" of Pub/Sub > settings. > > Sorry BTW, I should have brought this up a while ago, since we have so > many > > people using Salt now we are seeing a wide range of reports. > > Interesting! Are you hitting this purely on 2.x or have you seen it on > 3.x as well (and ideally master if tested)? Is there any consistency > in terms of how many connections the binding side has? And is there > any visible trigger event that seems to precede the stuck connections > appearing? > > Ian > _______________________________________________ > zeromq-dev mailing list > [email protected] > http://lists.zeromq.org/mailman/listinfo/zeromq-dev > I wish we had more visible triggers, it seems that it is a lot better when everything is the latest 2 series, but most of our user's deployments are still on 2. The Salt team are the guys that are pushing zeromq packages to a lot of distributions I should ask, should we be telling our package maintainers to distribute zeromq 3? All the packagers are still on 2
_______________________________________________ zeromq-dev mailing list [email protected] http://lists.zeromq.org/mailman/listinfo/zeromq-dev
