I was looking at support for cxgb*i, since open-iscsi previously only had DCB support for the software initiator. I'm guessing with cgdcbxd the task of looking up cgroup and assigning priority to the traffic is left to the individual transports now ? -Anish
One socket to bind them all. > -----Original Message----- > From: Mike Christie [mailto:micha...@cs.wisc.edu] > Sent: Monday, June 11, 2012 3:40 PM > To: open-iscsi@googlegroups.com > Cc: Anish Bhatt > Subject: Re: [PATCH] iscsid: Add support for net_prio cgroups > > On 06/08/2012 11:10 PM, Anish Bhatt wrote: > > Mike, > > Has there been any progress in figuring out how DCB would > > co-exist with net prio cgroups since you removed DCB support ? > > -Anish > > I think people were saying to do something with cgdcbxd. > > Are you guys using DCB with iSCSI for software iscsi or with cxgb*i? > > > > > > > On Monday, February 27, 2012 2:15:51 PM UTC-8, Rustad, Mark D wrote: > > > > Mike, > > > > On Feb 27, 2012, at 1:24 PM, Mike Christie wrote: > > > > > When it is changed, is there some event from the > kernel/userspace that > > > we can listen for or will it cause the current connection to be > come > > > disconnected and then on the reconnect we recheck and that is > how > > we get > > > the update? > > > > It is simpler and more direct than that. On packet egress, if the > > skb has no priority and came from a socket, the socket is checked > to > > see if it is in a cgroup. If it is, and a priority has been > > configured for that cgroup and interface, then the priority is > > applied to the skb at that point. So there is no session > disruption > > at all. > > > > A separate daemon takes DCB app priority changes as they happen > and > > sets them up for the necessary cgroups. So there is an event from > > the kernel, but this new daemon does all of the processing for > it. > > That is what drives the changes. > > > > Since if an skb already has a priority the net_prio cgroup code > will > > not be executed, this patch simply prevents iscsid from setting a > > priority when the net_prio cgroup exists. > > > > > Code wise patch looks ok > > > > Yeah, ultimately it would be nice to simply remove all of the > > DCB-related code from iscsid, since with the net_prio cgroups, it > > can all happen in a better way without any of it. The only reason > > not to do that now, is just to avoid any disruption to those > using > > it as it is today. I knew that the current DCB implementation > that I > > submitted last year would not stand - and I was working on a > better, > > simpler way to do it - but I was not expecting we could get to a > > place where no app change was needed at all. That was a very > > pleasant surprise. At least the current implementation enabled > some > > exploration and evaluation of iSCSI with DCB. I think you are in > a > > better position than I to judge when the DCB code can simply be > removed. > > > > -- > > Mark Rustad, LAN Access Division, Intel Corporation > > > > -- > > You received this message because you are subscribed to the Google > > Groups "open-iscsi" group. > > To view this discussion on the web visit > > https://groups.google.com/d/msg/open-iscsi/-/FmX53RYghW8J. > > To post to this group, send email to open-iscsi@googlegroups.com. > > To unsubscribe from this group, send email to > > open-iscsi+unsubscr...@googlegroups.com. > > For more options, visit this group at > > http://groups.google.com/group/open-iscsi?hl=en. -- You received this message because you are subscribed to the Google Groups "open-iscsi" group. To post to this group, send email to open-iscsi@googlegroups.com. To unsubscribe from this group, send email to open-iscsi+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/open-iscsi?hl=en.
