Why are there so many TIME_WAIT connections? #981
Comments
|
I have been investigating and when multiple incoming connections are being made, multiple connections to the backend services will be made. There are connection reuse strategies that some of the commercial load-balancers use, but that could also introduce issues with things like proxy protocol headers. If there was a large demand for implementing this feature it would be considered. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
I am using version 1.6.3 with HTTP proxy mode. Why does Fabio frequently create new connections to the same backend target service (e.g., 10.111.75.126) instead of reusing existing ones? When I check the Fabio machine(e.g.,10.111.76.65) using netstat, for example:
netstat -anp | grep '10.111.75.126' | grep 'TIME_WAIT'
I see around 200+ TIME_WAIT state entries. However, when I run:
netstat -anp | grep '10.111.75.126' | grep 'EST'
I only see 2 ESTABLISHED connections. Repeatedly executing this command shows that these ESTABLISHED ports change almost every second, indicating that connections are briefly used and then actively closed, resulting in a large number of TIME_WAIT states. If the request concurrency increases, won’t this lead to even more new connections and more TIME_WAIT entries? Is this a configuration issue?
------fabio.properties----
proxy.localip=127.0.0.1
proxy.strategy=rr
proxy.matcher=prefix
proxy.maxconn=10000
registry.consul.addr=10.111.76.65:8500
registry.consul.token=
registry.consul.kvpath=/fabio/config
registry.consul.tagprefix=urlprefix-
registry.consul.register.enabled = false
log.access.format = common
proxy.addr=:9999;proto=http
log.level = DEBUG
--------netstat----
[user@10_111_76_65 ~]# netstat -anp|grep '10.111.75.126'|grep 'TIME_WAIT' |wc -l
264
[user@10_111_76_65 ~]# netstat -anp|grep '10.111.75.126'|grep 'EST' |wc -l
2
The text was updated successfully, but these errors were encountered: