Hi, I am sorry I forgot to attach the necessary files :x.
On 09 Dec 2013, at 14:56, Lukas Tribus <luky...@hotmail.com> wrote: > Hi Annika, > > >> we have a few regarding load at our Haproxy 1.5-dev19 cluster. >> We run constantly at a load of 12 - 15 most of it is system load. >> [...] >> On our old cluster i do not see any of the "Resource temporarily >> unavailable” at splicing operation. > > We can't tell if that kind of load is normal for your box; please > don't make us guess from the context. > > Also please tell: > - hardware (cpu/ram/nic at least) on old/new cluster - Two Intel(R) Xeon(R) CPU X6550 @ 2.00GHz in each cluster node - 2x Emulex Corporation OneConnect 10Gb NIC (rev 02) in each cluster node - 32gbit RAM in each cluster node - Two nodes per cluster (active-active in the new one) > - software (kernel/OS) on old/new cluster - Debian Squeeze / 3.1.0-1-amd64 / Tickrate 250 - CentOS release 6.4 (Final) / 3.11.5-1.el6 / Tickrate 1000 > - HAProxy configuration on old/new cluster - Haproxy Configs use nbproc 24 without binding to specfic bind-process - Enabled SSL for 2 frontends on new cluster > - what is the actual number of concurrent sessions? - Cannot get the sessions from stats socket because of multiple processes - Firewall is seeing 300k concurrent sessions > > > >> Has something changed in kernel 3.11.5? > > Compared to what kernel release? > > > >> Are there any things which can be tried out at staging cluster >> to break down this problem? > > It seems you have a load problem; what happens if you disabling > splicing? - Going to try > > Are you using splice-auto or forcing splice by configuring > splice-request / splice-response? > - We are forcing by splice-request / splice-responce > > Could be a kernel thing, could be a NIC limitation or it could simply > be a higher load due to more concurrent connections … > - Concurrent connections staid the same during migration. The factor which we see is four times more load on the new loadbalancers. In our testing environment it does not seem to be such a hard impact. But the responsetimes stay the same regardless of the load. So is it just a displaying issue? And does not affect the users? Best regards, Annika > > > > Regards, > > Lukas
