Keepalive Timeout

[aliver]

Good day!

Our HTTP services often down because of high load.I want to increase keepalive timeout.

How possible to change keepalive timeout in services (not in script keepalive)?

I tried do this via keepalive frequency as saying in http://www.cisco.com/en/US/products/hw/contnetw/ps792/products_configuration_guide_chapter09186a008029c8be.html#wp1098307

But nothing changes and services still down.

Thanks.

[Gilles Dufour]

there are 2 timers.

The 'keepalive frequency' when the service is alive and the 'keepalive retryperiod' when the service is not alive.

So, you actually need to change both.

Regards,

Gilles.

Thanks for rating.

[aliver]

Hi Gilles,

well, I've change both (just not write about retryperiod in previous post)

Frequency - is a time between two keepalive sendings.

How timeout depends of frequency?

Now we have...

keepalive frequency 20

keepalive retryperiod 2

which timeout will be with these values?

Thanks.

[Gilles Dufour]

humm... with this setup, the interval between 2 keepalives is 2 sec [unless your forget a 0].

As a suggestion, always use the same value for both frequency and retryperiod.

The retryperiod is the interval between 2 keepalives when the service is not alive [as mentioned in my previous post].

Regards,

Gilles.

[aliver]

>>As a suggestion, always use the same value for both frequency and retryperiod

why? as i think the less retryperiod the less possible time of service down. thats why i choose 2 sec for it.

Let me ask otherwise...

How must i setup CSS if i want to have HTTP keepalive timeout >= 5 sec ?

Thanks.

[Gilles Dufour]

if the service is down because the server is overloaded, using a retryperiod of 2 sec will not help the server.

It means every 2 sec, the CSS is going to request the page.

Also, the connection timeout is deduced from the frequency/retryperiod.

So with 2sec, the timeout is 1sec.

This means your server has 1 sec to respond to the CSS.

Again if the server is overloaded, the chance to respond in 1sec is low.

So, use 20sec for both frequency and retryperiod.

Or 10sec.

Regards,

Gilles.

Thanks for rating.

[aliver]

I set 20sec to frequency and retryperiod.

So, situation is not change. Services continue down and quickly up many times. Intervals between down-up: from 1sec to 1min

[Gilles Dufour]

capture a sniffer trace and see what is going on.

My advice was just for how to set the timer correctly.

Without knowing exactly what is the issue - with sniffer trace - all you do is try and hope.

The first step towards resolution is always to fully understand the problem.

So, how many services do you have ?

How many are flapping ?

Sniffer trace ?

Was it working before or new install ?

Thanks,

Gilles.