-
questions about tuning a NUMA server
Hello everyone,
i have some questions about tuning q processes on NUMA server (linux kernel version is above 3.x).
1. According to https://code.kx.com/q/kb/linux-production/ [1] it is suggested to disable interleave and enable zone_reclaim, however https://code.kx.com/q/kb/cpu-affinity/#activating-numa [2] suggests
a) “numactl –interleave=all –physcpubind=0,1,2 q”
Q1. Why allows the process to access all node’s memory instead of bind the process to local memory only?
b) “echo 0 > /proc/sys/vm/zone_reclaim_mode” which lets the process closes zone-reclaim and takes other numa nodes’s memory.
Q2. Is zone_reclaim_mode=1 the correct answer according to [1] ?
The attached shows one of our q process, on a numa-enabled server, binds to mem 0-3,
Q3. Is it identical to “interleave=all” ?
Q4. If i want this process access local mem only, shall i execute “numactl –interleave=0” ?
Q5. following up Q4, How to start a q process and put numactl command’s to work in kx-platform? which Reversed Parameter shall be edited? or, should i “run”/start a process first, then execute system “numactl –interleave=0 –physcpubind=0 ” inside process’s console ?
thank you and the best,
leguan,
Log in to reply.