justin, yeah, this fabirc is all bran new mellanox, and all nodes are running their v2 stack. of for a beg report, sure thing. I was thinking i would tack on a comment here: https://bugzilla.redhat.com/show_bug.cgi?id=982757 since thats about the silent failure. -- Matthew Nicholson Research Computing Specialist Harvard FAS Research Computing matthew_nicholson at harvard.edu On Wed, Jul 10, 2013 at 3:00 PM, Justin Clift <jclift at redhat.com> wrote: > On 10/07/2013, at 7:49 PM, Matthew Nicholson wrote: > > Well, first of all,thank for the responses. The volume WAS failing over > the tcp just as predicted,though WHY is unclear as the fabric is know > working (has about 28K compute cores on it all doing heavy MPI testing on > it), and the OFED/verbs stack is consistent across all client/storage > systems (actually, the OS image is identical). > > > > Thats quiet sad RDMA isn't going to make 3.4. We put a good deal of > hopes and effort around planning for 3.4 for this storage systems, > specifically for RDMA support (well, with warnings to the team that it > wasn't in/test for 3.3 and that all we could do was HOPE it was in 3.4 and > in time for when we want to go live). we're getting "okay" performance out > of IPoIB right now, and our bottle neck actually seems to be the fabric > design/layout, as we're peaking at about 4.2GB/s writing 10TB over 160 > threads to this distributed volume. > > Out of curiosity, are you running the stock OS provided infiniband stack, > or are you using the "vendor optimised" version? (eg "Mellanox OFED" if > you're using Mellanox cards) > > Asking because although I've not personally done any perf measurements > between them, Mellanox swears the new v2 of their OFED stack is much higher > performance than both the stock drivers or their v1 stack. IPoIB is > especially tuned. > > I'd really like to get around to testing that some time, but it won't be > soon. :( > > > > When it IS ready and in 3.4.1 (hopefully!), having good docs around it, > and maybe even a simple printf for the tcp failover would be huge for us. > > Would you be ok to create a Bugzilla ticket, asking for that printf item? > > > https://bugzilla.redhat.com/enter_bug.cgi?product=GlusterFS&component=rdma > > It doesn't have to be complicated or super in depth or anything. :) > > Asking because when something is a ticket, the "task" is much easier to > hand > to someone so it gets done. > > If that's too much effort though, just tell me what you'd like as the > ticket > summary line + body text and I'll go create it. :) > > Regards and best wishes, > > Justin Clift > > -- > Open Source and Standards @ Red Hat > > twitter.com/realjustinclift > > -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://supercolony.gluster.org/pipermail/gluster-users/attachments/20130710/32247700/attachment.html>