deploy/cobbler: drop tcp_tw_recycle in sysctl.conf 21/49521/1
authorYibo Cai <yibo.cai@arm.com>
Thu, 21 Dec 2017 09:20:09 +0000 (17:20 +0800)
committerYibo Cai <yibo.cai@arm.com>
Fri, 22 Dec 2017 02:17:44 +0000 (10:17 +0800)
Currently, tcp_tw_recycle is set explicitly in sysctl.conf when
installing OS by cobbler. This option is removed in kernel4.12,
current sysctl.conf breaks sysctl function and K8s deployment.

Actually, tcp_tw_recycle is considered broken and not recommended to
use. We may simply drop this option to be compatible with different
kernel versions.

References:
- https://vincent.bernat.im/en/blog/2014-tcp-time-wait-state-linux
- https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=4396e46187ca5070219b81773c4e65088dac50cc

Change-Id: I2cd89db3d423c811ba58eb5960f3cc718475be5f
Signed-off-by: Yibo Cai <yibo.cai@arm.com>
deploy/adapters/cobbler/snippets/kickstart_sysctl.conf
deploy/adapters/cobbler/snippets/preseed_sysctl.conf
deploy/adapters/cobbler/snippets/sysctl.xml

index c227ecf..112f010 100644 (file)
@@ -59,7 +59,6 @@ net.ipv4.ip_local_port_range = 15000 61000
 net.ipv4.tcp_fin_timeout=30
 
 # fast cycling of sockets in time_wait state and re-using them
-net.ipv4.tcp_tw_recycle = 1
 net.ipv4.tcp_tw_reuse = 1
 
 # increase the maximum number of requests queued to a listen socket
index c227ecf..112f010 100644 (file)
@@ -59,7 +59,6 @@ net.ipv4.ip_local_port_range = 15000 61000
 net.ipv4.tcp_fin_timeout=30
 
 # fast cycling of sockets in time_wait state and re-using them
-net.ipv4.tcp_tw_recycle = 1
 net.ipv4.tcp_tw_reuse = 1
 
 # increase the maximum number of requests queued to a listen socket
index fe13bf7..42f9597 100644 (file)
@@ -62,7 +62,6 @@ net.ipv4.ip_local_port_range = 15000 61000
 net.ipv4.tcp_fin_timeout=30
 
 # fast cycling of sockets in time_wait state and re-using them
-net.ipv4.tcp_tw_recycle = 1
 net.ipv4.tcp_tw_reuse = 1
 
 # increase the maximum number of requests queued to a listen socket