redis.conf详解之tcp-keepalive

用法

单位秒

tcp-keepalive 300

 

用途

在linux系统中，客户端发送的最后一个数据包与redis发送的第一个保活探测报文之间的时间间隔。单位是秒。
 

注意事项

1.tcp-keepalive只在linux生效，在其他系统不生效(比如mac)，只按照其他系统内核自身的设置。
2.keepalive在linux系统中只控制tcp_keepalive_time选项。

 

linux保活定时器相关参数

cat /proc/sys/net/ipv4/tcp_keepalive_time 
cat /proc/sys/net/ipv4/tcp_keepalive_intvl KeepAlive探测包的发送间隔，默认值为75s
cat /proc/sys/net/ipv4/tcp_keepalive_probes 收不到响应，共计发几次。

 

mac系统测试

结论：与是否设置tcp-keepalive毫无关系

设置mac内核参数net.inet.tcp.keepidle为3秒

sudo sysctl -w net.inet.tcp.keepidle=3000

redis获取tcp-keepalive配置（300s），并不在发送任何请求。

127.0.0.1:6379> config get tcp-keepalive
1) "tcp-keepalive"
2) "300"

观察tcp情况，发现每隔3～4秒探测一次（客户端未关闭，所以每次都得到响应）,tcp-keepalive未生效。

$ sudo tcpdump -i lo0 port 6379

18:07:24.933364 IP localhost.6379 > localhost.63787: Flags [.], ack 1, win 6378, length 0
18:07:24.933382 IP localhost.63787 > localhost.6379: Flags [.], ack 1, win 6378, length 0
18:07:24.933427 IP localhost.63787 > localhost.6379: Flags [.], ack 1, win 6378, options [nop,nop,TS val 2931790987 ecr 2245109217], length 0
18:07:24.933436 IP localhost.6379 > localhost.63787: Flags [.], ack 1, win 6378, options [nop,nop,TS val 2245112230 ecr 2931787974], length 0

18:07:28.040568 IP localhost.6379 > localhost.63787: Flags [.], ack 1, win 6378, length 0
18:07:28.040577 IP localhost.63787 > localhost.6379: Flags [.], ack 1, win 6378, length 0
18:07:28.040592 IP localhost.63787 > localhost.6379: Flags [.], ack 1, win 6378, options [nop,nop,TS val 2931794047 ecr 2245112230], length 0
18:07:28.040596 IP localhost.6379 > localhost.63787: Flags [.], ack 1, win 6378, options [nop,nop,TS val 2245115290 ecr 2931790987], length 0

18:07:31.105544 IP localhost.6379 > localhost.63787: Flags [.], ack 1, win 6378, length 0
18:07:31.105571 IP localhost.63787 > localhost.6379: Flags [.], ack 1, win 6378, length 0
18:07:31.105592 IP localhost.63787 > localhost.6379: Flags [.], ack 1, win 6378, options [nop,nop,TS val 2931797050 ecr 2245115290], length 0
18:07:31.105603 IP localhost.6379 > localhost.63787: Flags [.], ack 1, win 6378, options [nop,nop,TS val 2245118293 ecr 2931794047], length 0

 

linux系统测试

结论：redis的tcp-keepalive覆盖了linux的tcp_keepalive_time（默认是2小时）

redis获取tcp-keepalive配置（6s），并不在发送任何请求。

127.0.0.1:6379> config get tcp-keepalive
1) "tcp-keepalive"
2) "6"

观察tcp情况，发现每隔6秒探测一次（客户端未关闭，所以每次都得到响应）,tcp-keepalive生效了！

$ sudo tcpdump -i lo0 port 6379
13:50:42.731182 IP localhost.41692 > localhost.redis: Flags [P.], seq 18:63, ack 11469, win 512, options [nop,nop,TS val 4084903638 ecr 4084900541], length 45: RESP "config" "get" "tcp-keepalive"
13:50:42.731243 IP localhost.redis > localhost.41692: Flags [.], ack 63, win 512, options [nop,nop,TS val 4084903638 ecr 4084903638], length 0
13:50:42.731471 IP localhost.redis > localhost.41692: Flags [P.], seq 11469:11500, ack 63, win 512, options [nop,nop,TS val 4084903638 ecr 4084903638], length 31: RESP "tcp-keepalive" "6"
13:50:42.731538 IP localhost.41692 > localhost.redis: Flags [.], ack 11500, win 512, options [nop,nop,TS val 4084903638 ecr 4084903638], length 0

13:50:48.759844 IP localhost.redis > localhost.41692: Flags [.], ack 63, win 512, options [nop,nop,TS val 4084909667 ecr 4084903638], length 0
13:50:48.759896 IP localhost.41692 > localhost.redis: Flags [.], ack 11500, win 512, options [nop,nop,TS val 4084909667 ecr 4084903638], length 0

13:50:54.742957 IP localhost.redis > localhost.41692: Flags [.], ack 63, win 512, options [nop,nop,TS val 4084915684 ecr 4084909667], length 0
13:50:54.743012 IP localhost.41692 > localhost.redis: Flags [.], ack 11500, win 512, options [nop,nop,TS val 4084915684 ecr 4084903638], length 0

13:50:57.748679 IP localhost.41692 > localhost.redis: Flags [.], ack 11500, win 512, options [nop,nop,TS val 4084918690 ecr 4084903638], length 0
13:50:57.748720 IP localhost.redis > localhost.41692: Flags [.], ack 63, win 512, options [nop,nop,TS val 4084918690 ecr 4084915684], length 0

 

原生注释

# TCP keepalive.
#
# If non-zero, use SO_KEEPALIVE to send TCP ACKs to clients in absence
# of communication. This is useful for two reasons:
#
# 1) Detect dead peers.
# 2) Force network equipment in the middle to consider the connection to be
#    alive.
#
# On Linux, the specified value (in seconds) is the period used to send ACKs.
# Note that to close the connection the double of the time is needed.
# On other kernels the period depends on the kernel configuration.
#
# A reasonable value for this option is 300 seconds, which is the new
# Redis default starting with Redis 3.2.1.
#tcp-keepalive 300