>> On Tue, Mar 1, 2016 at 1:27 PM, Amedeo Perazzo >> > wrote: >> >> I noticed we lost again the network optimization settings on the >> psexport nodes. I propose we apply the following settings to all of >> our nodes, diskless and non-diskless. At the end, I add some >> settings which should be applied only to psexport, if possible. >> >> Thank you for looking at this, >> Amedeo >> -- >> >> All nodes: >> ---------- >> >> # Controls the minimum amount of free memory, helps with OOM >> vm.min_free_kbytes = 131072 >> >> # Network buffer sizes 32M >> net.core.rmem_max = 33554432 >> net.core.wmem_max = 33554432 >> >> >> DAQ nodes and analysis >> ---------------------- >> >> (These are all the daq-* - mainly diskless, these should already be >> done, just check) >> >> # Network buffer sizes 128M >> net.core.rmem_max = 134217728 >> net.core.wmem_max = 134217728 >> >> >> psexport nodes >> -------------- >> >> # Network buffer sizes 128M >> net.core.rmem_max = 134217728 >> net.core.wmem_max = 134217728 >> >> # increase Linux autotuning TCP buffer limit to 32MB >> net.ipv4.tcp_rmem = 4096 87380 33554432 >> net.ipv4.tcp_wmem = 4096 65536 33554432 >> >> # increase the length of the processor input queue >> net.core.netdev_max_backlog = 30000 >> >> # recommended default congestion control is htcp >> net.ipv4.tcp_congestion_control=htcp >> >> # recommended for hosts with jumbo frames enabled >> net.ipv4.tcp_mtu_probing=1 >> >> ######################################################################## >> Use REPLY-ALL to reply to list