Just to follow up on this, the packet loss appeared to be an unrelated issue. If I connect directly to a realserver, the page loads fairly fast. If I connect through the loadbalancer ( and get the same real server) I see a lot of tcp retranmissions and such (as sniffed from the real server). I can't figure it out. The delays appear to be from the RTO (retrans timeout). I saw a patch from miccy soft about the mtu discovery and have installed, I also tried disabling mtu discovery.
Is lvs-nat "more compatible"? On 8/2/07, Dan Baughman <[EMAIL PROTECTED]> wrote: > > Well we are getting 1 - 5% packet loss to one of the real servers. We may > have a layer 2 issue here, or perhaps something physical I'll let you know > what we turn up. Thanks for your time and suggestions. > > > On 8/1/07, Joseph Mack NA3T <[EMAIL PROTECTED]> wrote: > > > > On Wed, 1 Aug 2007, Dan Baughman wrote: > > > > > Thanks for the suggestions so far, here is what the dumps yielded. > > > > > > I'm using ethereal to look at a side-by-side comparison of > > > a healthy conversation and one of the delayed > > > conversations. The delayed conversation has two tcp > > > packet types that the normal conversation doesn't: > > > > > > 1) A lot of duplicate acks from the client to the server > > > > do both get to the realserver? > > > > > 2) TCP Retransmissions from the server to client. > > > > do both transmissions arrive at the client? (I assume they > > leave the realserver.) > > > > > It seems that there is a lot of traffic being duplicated > > > and received by the real server twice. This isn't > > > occuring when they access the realserver ip to view the > > > site. The extra load time is occuring after the server > > > receives such a duplicate ack, its a pause of about 3 > > > seconds, then a tcp retransmission occurs. > > > > hmm. Nothing springs to mind. > > > > do you have any iptables rules anywhere? > > > > is the unhealthy connection from clients all at one > > site/coming in from one provider? > > > > > I guess it might be a version issue, 'ivsadm -v' yields: > > > ipvsadm v1.24 2003/06/07 (compiled with popt and IPVS > > > v1.2.0) > > > > do you have a matching ipvsadm/ip_vs pair? > > > > Joe > > > > -- > > Joseph Mack NA3T EME(B,D), FM05lw North Carolina > > jmack (at) wm7d (dot) net - azimuthal equidistant map > > generator at http://www.wm7d.net/azproj.shtml > > Homepage http://www.austintek.com/ It's GNU/Linux! > > > > _______________________________________________ > > LinuxVirtualServer.org mailing list - [email protected] > > Send requests to [EMAIL PROTECTED] > > or go to http://lists.graemef.net/mailman/listinfo/lvs-users > > > > _______________________________________________ LinuxVirtualServer.org mailing list - [email protected] Send requests to [EMAIL PROTECTED] or go to http://lists.graemef.net/mailman/listinfo/lvs-users
