像石头一样翻滚吧
发现用户喜欢用while … sleep…来写Dockerfile的 ENTRYPOINT, 忍不住思考，是tail -f /dev/null的开销小,还是while true
do
sleep 3
done 更小？
tail -f /dev/null is better，因为tail -f 会一直等待 I/O 完成，进程一直处于Waiting状态。
而 while 每隔3秒会重新进行一次进程上下文切换，sleep 是单独起1个进程，而sleep结束后，会通过中断通知cpu，通过strace可以跟踪两者的系统调用
1 | #strace tail -f /dev/null |
在 read(4 卡住。。。 因为I/O 无法完成读取/dev/null
1 | # strace -p PID_OF_WHILE_SLEEP_SHELL |
SELECT * FROM k8s.”default”.”cpu/node_utilization” WHERE (nodename =~ /^10.35.48.177$/ AND type = ‘node’) AND time >= now() - 1h;
或者
SELECT mean(value) FROM k8s.”default”.”cpu/node_utilization” WHERE (nodename =~ /^10.35.48.177$/ AND type = ‘node’) AND time >= now() - 1h GROUP BY time(1m)
SELECT mean(value) / 1000 FROM k8s.”default”.”cpu/node_capacity” WHERE (type = ‘node’ AND nodename =~ /^10.35.48.177$/) AND time >= now() - 5m GROUP BY time(30s)
部分sql语法可以自己斟酌下哈
每1分钟聚合的内存利用率均值: SELECT mean(value) FROM k8s."default"."memory/node_utilization" WHERE (nodename =~ /^10\.35\.48\.177$/ AND type = 'node') AND time >= now() - 1h GROUP BY time(1m)
SELECT mean(value) / 1000 FROM k8s.”default”.”cpu/node_capacity” WHERE (type = ‘node’ AND nodename =~ /^10.35.48.177$/) AND time >= now() - 1h GROUP BY time(30s)
SELECT mean(value) FROM k8s.”default”.”memory/node_allocatable” WHERE (type = ‘node’ AND nodename =~ /^10.35.48.177$/) AND time >= now() - 1h GROUP BY time(30s)
SELECT value FROM k8s.”default”.”memory/usage” WHERE (type = ‘node’ AND nodename =~ /^10.35.48.177$/) AND time >= now() - 1h
SELECT value FROM k8s.”default”.”memory/rss” WHERE (type = ‘node’ AND nodename =~ /^10.35.48.177$/) AND time >= now() - 1h
SELECT value FROM k8s.”default”.”memory/cache” WHERE (type = ‘node’ AND nodename =~ /^10.35.48.177$/) AND time >= now() - 1h
SELECT value FROM k8s.”default”.”memory/working_set” WHERE (type = ‘node’ AND nodename =~ /^10.35.48.177$/) AND time >= now() - 1h
SELECT value FROM k8s.”default”.”network/rx_rate” WHERE (type = ‘node’ AND nodename =~ /^10.35.48.177$/) AND time >= now() - 1h
SELECT value FROM k8s.”default”.”network/tx_rate” WHERE (type = ‘node’ AND nodename =~ /^10.35.48.177$/) AND time >= now() - 1h
SELECT mean(value) FROM k8s.”default”.”disk/io_read_bytes_rate” WHERE (type = ‘node’ AND nodename =~ /^10.35.48.177$/) AND time >= now() - 1h GROUP BY time(30s)
SELECT mean(value) FROM k8s.”default”.”disk/io_write_bytes_rate” WHERE (type = ‘node’ AND nodename =~ /^10.35.48.177$/) AND time >= now() - 1h GROUP BY time(30s)
SELECT value FROM k8s.”default”.”filesystem/available” WHERE (type = ‘node’ AND nodename =~ /^10.35.48.177$/) AND time >= now() - 10m group by resource_id
SELECT value FROM k8s.”default”.”filesystem/usage” WHERE (type = ‘node’ AND nodename =~ /^10.35.48.177$/) AND time >= now() - 10m group by resource_id
SELECT value FROM k8s.”default”.”filesystem/limit” WHERE (type = ‘node’ AND nodename =~ /^10.35.48.177$/) AND time >= now() - 1m group by resource_id
SELECT mean(value) FROM k8s.”default”.”accelerator/memory_total” WHERE nodename =~ /^10.30.21.153$/ AND time >= now() - 1h GROUP BY time(1m)
SELECT mean(value) FROM k8s.”default”.”accelerator/memory_used” WHERE nodename =~ /^10.30.21.153$/ AND time >= now() - 1h GROUP BY time(1m)
SELECT value FROM k8s.”default”.”accelerator/duty_cycle” WHERE (type = ‘node’ AND nodename =~ /^10.35.48.177$/) AND time >= now() - 10m
SELECT value FROM k8s.”default”.uptime WHERE (type = ‘node’ AND nodename =~ /^10.35.48.177$/) AND time >= now() - 1m
对于多副本(pod)的无状态服务/有状态服务/短任务，查询到的指标值应该是各个副本的和
1 | SELECT mean(value) FROM k8s."default"."memory/usage" WHERE (type = 'rc' AND rc_name =~ /^autotest-template-server123-4$/ AND namespace_name =~ /^guxiawei-84$/) AND time >= now() - 1h GROUP BY time(1m) |
1 | SELECT mean(value) FROM k8s."default"."memory/usage" WHERE (type = 'job' AND job_name =~ /^job-createzy6vgx$/ AND namespace_name =~ /^liuguiying-61$/) AND time >= now() - 1h GROUP BY time(1m) |
1 | SELECT mean(value) FROM k8s."default"."memory/usage" WHERE (type = 'ss' AND ss_name =~ /^testservice110$/ AND namespace_name =~ /^wangmeng3-154$/) AND time >= now() - 1h GROUP BY time(1m) |
1 | SELECT mean(value) FROM k8s."default"."memory/usage" WHERE (type = 'pod' AND pod_name =~ /^testservice110-0$/ AND namespace_name =~ /^wangmeng3-154$/) AND time >= now() - 1h GROUP BY time(1m) |
根据pod名称和namespace查询，并按照容器名分类
1 | SELECT value FROM k8s."default"."memory/usage" WHERE (type='pod_container' AND pod_name='gb-groupdn-stable-779-1-rcv9x' AND namespace_name =~ /^guanbo-225$/) AND time >= now() - 5m group by container_name |
按照容器名和pod名和namespace查询
1 | SELECT value FROM k8s."default"."memory/usage" WHERE (type='pod_container' AND pod_name='gb-groupdn-stable-779-1-rcv9x' AND namespace_name =~ /^guanbo-225$/ and container_name='media-node') AND time >= now() - 5m |
按照容器名和namespace查询（不推荐，容器名可能重名）
1 | SELECT value FROM k8s."default"."memory/usage" WHERE (type='pod_container' AND namespace_name =~ /^guanbo-225$/ and container_name='media-node') AND time >= now() - 5m |
1 | SELECT mean(value) FROM k8s."default"."accelerator/memory_used" WHERE (type = 'pod' AND pod_name =~ /^autotest-template-server123-4-87ms1$/ AND namespace_name =~ /^haoyuan-8$/) AND time >= now() - 1h GROUP BY time(1m) |
对于容器，需要的指标有：
1 | cpu/limit |
1 | authorizer, err := apiserver.NewAuthorizerFromAuthorizationConfig(s.AuthorizationMode, s.AuthorizationPolicyFile) |
static void
putch(int ch, int cnt)
{
cputchar(ch);
cnt++;
}
int
vcprintf(const char *fmt, va_list ap)
{
int cnt = 0;
vprintfmt((void*)putch, &cnt, fmt, ap);
return cnt;
}
int
cprintf(const char *fmt, …)
{
va_list ap;
int cnt;
va_start(ap, fmt);
cnt = vcprintf(fmt, ap);
va_end(ap);
return cnt;
}
重新开始刷清华ucore , 取材自MIT 6.828, 操作系统神课，把类似unix的内核手动实现一遍。周末刷了物理内存管理（双向链表维护空闲内存，多级页表映射）
目标今年向向docker贡献源代码！往kubernetes提交pr！先熟悉内核。