something that counted the number of outstanding connections by
returning a wrapped net.Conn where you hook the Close() method.
Elevated CLOSE_WAITs _may_ indicate that the server is not responding
to the clients' close requests in time, just a guess. You could also
be bashing your head against some ec2 load balancer problem, also a
guess.
On Sat, Nov 2, 2013 at 6:41 PM, Karim Nassar wrote:
Sure. go version go1.1.2 linux/amd64, running debian wheezy 7.2 on google
compute "high cpu" instances (8 cpu, 7.2GB memory)
--
Karim Nassar
Software Engineer, eclipse.io
(310) 598-0003
--Sure. go version go1.1.2 linux/amd64, running debian wheezy 7.2 on google
compute "high cpu" instances (8 cpu, 7.2GB memory)
--
Karim Nassar
Software Engineer, eclipse.io
(310) 598-0003
On Sat, Nov 2, 2013 at 12:37 AM, Dave Cheney wrote:
Can you give a few more details about your environment ? Ie, which os
(i'm guessing linux inside ec2), which version of go, etc.
Can you give a few more details about your environment ? Ie, which os
(i'm guessing linux inside ec2), which version of go, etc.
On Sat, Nov 2, 2013 at 6:24 PM, wrote:
I am working on a high volume (7-10k reqs/sec) http application that is
at
times getting overwhelmed with occasional traffic spikes during peak
traffic
times. When this happens, the number of CLOSE_WAIT connections grows
into
the 10s of thousands, Reducing tcp kernel timeouts helps recovery, but
right now I am totally reactive and using linux command line tools to
monitor. I would like to detect this situation in the application where
I
can remediate. Is there a way to hook into the tcp state transitions in
the
app? RIght now I start a http ServerMux using ListenAndServe(). Do I
need to
roll my own? Any tips or links to docs would be appreciated.
Thanks in advance!
Karim
--
You received this message because you are subscribed to the Google
Groups
"golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send
an
email to golang-nuts+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
I am working on a high volume (7-10k reqs/sec) http application that is
at
times getting overwhelmed with occasional traffic spikes during peak
traffic
times. When this happens, the number of CLOSE_WAIT connections grows
into
the 10s of thousands, Reducing tcp kernel timeouts helps recovery, but
right now I am totally reactive and using linux command line tools to
monitor. I would like to detect this situation in the application where
I
can remediate. Is there a way to hook into the tcp state transitions in
the
app? RIght now I start a http ServerMux using ListenAndServe(). Do I
need to
roll my own? Any tips or links to docs would be appreciated.
Thanks in advance!
Karim
--
You received this message because you are subscribed to the Google
Groups
"golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send
an
email to golang-nuts+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
You received this message because you are subscribed to the Google Groups "golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
