bpftune - BPF driven auto-tuning | install & usage
1. First | 序
bpftune 是由 Oracle 开源的一款基于 eBPF（Extended Berkeley Packet Filter） 技术的高级 Linux 内核自动调优工具。它的核心理念是提供一种轻量级、常驻（Always-on）、自适应的系统行为优化方案。
此篇记录安装&使用过程
📦
本篇所使用到的操作系统&软件版本(配置)如下所示
2. Install | 安装
-
使用 transactional-update 安装 bpftune
transactional-update pkg install bpftune -
Reboot after update 更新后重启
transactional-update reboot -
重启完成后启用 bpftune
systemctl enable --now bpftune.service
3. Usage | 使用
systemctl enable --now bpftune.service 开启相应服务后 bpftune 立即开始运行
bpftune 命令一些帮助文档
~> bpftune --help
Usage: /sbin/bpftune [OPTIONS]
OPTIONS := { { -a|--allow tuner}
{ -d|--debug} {-D|--daemon}
{ -c|--cgroup cgroup_path}
{ -L|--legacy}
{ -h|--help}}
{ -l|--libdir library_path}
{ -p|--port port}
{ -q|--query query}
{ -r|--learning_rate learning_rate}
{ -R|--rollback}
{ -s|--stderr}
{ -S|--suppport}
{ -V|--version}}
~> bpftune -q help
help list supported queries
tuners show state of tuners
tunables show list of tunables
jtunables show list of tunables(json)
summary show summary of changes
status show status of tunables
jstatus show status of tunables(json)
rollback show changes needed to roll back
-
查看系统是否支持 bpftune
~> bpftune -S bpftune: bpftune works fully bpftune: bpftune supports per-netns policy (via netns cookie) -
查看 tuners 状态
~> bpftune -q tuners ip_frag active neigh_table active net_buffer active netns active route_table active sysctl active tcp_buffer active tcp_conn active udp_buffer active -
显示可调参数状态
~> bpftune -q status ip_frag net.ipv4.ipfrag_high_thresh 4194304 ip_frag net.ipv6.ip6frag_high_thresh 4194304 neigh_table net.ipv4.neigh.default.gc_interval 30 neigh_table net.ipv4.neigh.default.gc_stale_time 60 neigh_table net.ipv4.neigh.default.gc_thresh1 128 neigh_table net.ipv4.neigh.default.gc_thresh2 512 neigh_table net.ipv4.neigh.default.gc_thresh3 1024 neigh_table net.ipv6.neigh.default.gc_interval 30 neigh_table net.ipv6.neigh.default.gc_stale_time 60 neigh_table net.ipv6.neigh.default.gc_thresh1 128 neigh_table net.ipv6.neigh.default.gc_thresh2 512 neigh_table net.ipv6.neigh.default.gc_thresh3 1024 net_buffer net.core.netdev_max_backlog 1000 net_buffer net.core.flow_limit_cpu_bitmap 0 net_buffer net.core.netdev_budget 300 net_buffer net.core.netdev_budget_usecs 2000 net_buffer net.core.high_order_alloc_disable 0 route_table net.ipv6.route.max_size 2147483647 tcp_buffer net.ipv4.tcp_wmem 4096 16384 4194304 tcp_buffer net.ipv4.tcp_rmem 4096 131072 33554432 tcp_buffer net.ipv4.tcp_mem 187011 249351 374022 tcp_buffer net.ipv4.tcp_moderate_rcvbuf 1 tcp_buffer net.ipv4.tcp_no_metrics_save 0 tcp_buffer net.ipv4.tcp_no_ssthresh_metrics_save 1 tcp_buffer net.core.high_order_alloc_disable 0 tcp_buffer net.ipv4.tcp_max_syn_backlog 1024 tcp_buffer net.ipv4.tcp_max_orphans 65536 tcp_conn net.ipv4.tcp_allowed_congestion_control "reno cubic bbr htcp dctcp" tcp_conn net.ipv4.tcp_available_congestion_control "reno cubic bbr htcp dctcp" tcp_conn net.ipv4.tcp_congestion_control "cubic" tcp_conn net.ipv4.tcp_thin_linear_timeouts 0 udp_buffer net.ipv4.udp_mem 374025 498702 748050 udp_buffer net.core.rmem_max 4194304 udp_buffer net.core.rmem_default 212992
Thanks & Copy
GitHub - oracle/bpftune: bpftune uses BPF to auto-tune Linux systems
bpftune uses BPF to auto-tune Linux systems. Contribute to oracle/bpftune development by creating an account on GitHub.
oracle
