Web Monitoring response time problems

[robson]

Hello everyone,

We are testing the Zabbix solution at our network and specifically we are using the web monitoring feature to monitor some of our web services.
After adding some scenarios with their respective steps, everything went well until we realized there was something weird regarding the response time.

As you can see at the image, there are more than five seconds peaks. And it's always more than five seconds, it's never less than that. It's like a repeated pattern.

We have checked the curl requests through the tcpdump in the zabbix server and there is no such delay of five seconds. There is no error messages in zabbix_server.log as well.


Any clues ?

Zabbix version: 2.0.2
Ubuntu server version 12.04.1 LTS

[robson]

The reason

I've finally found out the reason for that.

After a lot of research on the internet and studying the http poller behaviour, the reason for those five seconds delays is the DNS connection used by the poller in order to resolve the web server IP address.

By analysing the hole tcpdump packets through net interface, I realized that http poller usually sends 3 DNS requests on sequence everytime it needs to resolve an IP address. But sometimes, the http poller sends ONLY one DNS request. I don't know why.

Our internal DNS only responds to more than one request. If there is only one, it doesn't respond to it. I don't know why as well.

Especially in that case, when the http poller sends only one request and our internal DNS doesn't respond to it, it forces it to wait for the response in 5 seconds. After the 5 seconds, the poller sends the DNS request again but with 3 on sequence being succesful in resolving the IP address and finally connecting to the web server.

For now, I have fixed the problem by editing /etc/hosts where I set the DNS web server names and their respective IP addresses so that http poller doesn't need to resolve them.

[jnsvano]

Hi,

I'm new to Zabbix and having the same problem. Monitored web page returns about 200-300b of data. When measure response time through browser, it's always under 50ms, but Zabbix (2.0.2 on Ubuntu 12.04) always shows more than 200ms, sometimes seconds.
This is a simple scenario with only one step (Required string: "OK", Required status codes: "200").
Tried the suggested solution: added web host to /etc/hosts, but nothing changed at all. Saw the zabbix_server.log, but nothing there about it.