timeout in step of web scenario don't work

[trible]

Hi,

I have problems with timeout definition in web scenario :
whatever time I'am inputing in "timeout", it's update interval that define status of scenario.

For example, with 15s of timeout in step one of my scenario and 35s of update interval, when I stop services of web server for 5s just at the start of update, the scenario don't wait the end of timeout and status is off until next update.

If you can help me ...

best regards.

[Aly]

Timeout is for receiving data, not for connection estimation.

[trible]

Timeout is for receiving data, not for connection estimation.

Thank you Aly. I can believe what is Timeout but I badly expressed myself.

My scenario want read an expression on my home page. If the timeout is 15s and if the home page is unavailable during 5s, I shouldn't have a discontinuity of the web scenario's graph because timeout is condition for result of scenario (like "riquired" "URL" or "Status Code"). Am I right ?

[hardtofi]

Thank you Aly. I can believe what is Timeout but I badly expressed myself.

My scenario want read an expression on my home page. If the timeout is 15s and if the home page is unavailable during 5s, I shouldn't have a discontinuity of the web scenario's graph because timeout is condition for result of scenario (like "riquired" "URL" or "Status Code"). Am I right ?

I would say you're not. You don't get a timeout if the web server is down, you get a connection refused. Any browser will at that point give up which means down time, not slow web server time.

[trible]

I would say you're not. You don't get a timeout if the web server is down, you get a connection refused. Any browser will at that point give up which means down time, not slow web server time.

Thank's for reply hardtofi. I see I was mistaken about timeout. However, on production's servers, I can note timout isn't respected. Even the web server is all right, remote zabbix server see its down 3 or 4 times per hour (because a response time is sometimes larger than 5 seconds and if this response time fall in update's scenario moment, result isn't what I would like).

So I think the timeout option don't working correctly ?

Am I alone in this case ?

[hardtofi]

Thank's for reply hardtofi. I see I was mistaken about timeout. However, on production's servers, I can note timout isn't respected. Even the web server is all right, remote zabbix server see its down 3 or 4 times per hour (because a response time is sometimes larger than 5 seconds and if this response time fall in update's scenario moment, result isn't what I would like).

So I think the timeout option don't working correctly ?

Am I alone in this case ?

The timeout option is the point where you want the web service to give up. That means it fails, the same with the matching regexp or response code you can set. If any of these requirements fail it won't be plotted but will only be considered a failure.

What I first tried with was having a large timeout and then making "IT-services" http://www.zabbix.com/forum/showthread.php?t=8920 using triggers to see when the SLA was broken. However IT-Services seems very broken so that was a no go.

Instead I've done like this:
Define a web check with a 32s timout. If it takes longer than that I consider the request a total failure. Then I have three triggers:
If web request fails or resp time > 30s - critical
If resp time > 5s (depending on above trigger 30s trigger) - warning
If resp time > 1s (depending on above 5s trigger) - info

Then I can see under reports how many percentage of the requests takes over 1, 5 and 30s and I can also see under monitoring->web the avg times.

This is a very clunky way of doing it but I haven't unfortunately found a simple way that actually works.