Discussion thread for official Zabbix Template HAProxy

[AlexL]

This thread is designed to provide grounds for discussion of the official Zabbix Template for HAProxy.
The template and details of the template are available in GIT repository.
HAProxy Agent https://git.zabbix.com/projects/ZBX/.../haproxy_agent
HAProxy HTTP https://git.zabbix.com/projects/ZBX/...p/haproxy_http
Zabbix is always looking for ways to improve our services and to make our users happier.
We pride ourselves on doing our best each and every day, but we know that there is always something more to learn.
We would like to hear back from you to know what have you liked and what would you improve in the template.

[rstumbaum]

Hi,

and thanks for providing this official HAProxy template!

My use case:
I have some haproxy services running on active-passive heartbeat loadbalancers.

When using the haproxy_agent version I get the problem that the inactive host is complaining that the HAProxy service is not running - since it gets started only on failover by heartbeat.

When using the haproxy_http on the virtual IP I get the problem that I have to maintain for each Virtual-IP Host entry in Zabbix the {$HAPROXY.STATS.HOST} macro.


Suggestion:
Could you replace the {$HAPROXY.STATS.HOST} in the item configurations to {HOST.CONN} instead as it is in the haproxy_agent version?

Best regards and a Happy New Year!
Rainer

[max.ch.88]

Hi,

Suggestion:
Could you replace the {$HAPROXY.STATS.HOST} in the item configurations to {HOST.CONN} instead as it is in the haproxy_agent version?

https://git.zabbix.com/projects/ZBX/...Frelease%2F4.4
The template is using {HOST.CONN} now.

[rstumbaum]

There are two templates: agent based and http based. I have been writing about the http based one - you looked at the agent based one, where it is correct.

It should be fixed here: https://git.zabbix.com/projects/ZBX/...Frelease%2F4.4

[max.ch.88]

Both templates have the same triggers. And the "HAProxy service is not running" trigger will be lit even if you change {$HAPROXY.STATS.HOST} to {HOST.CONN}.
Is it a big problem to maintain macro?
An agentless template can be attached to a host with the agent checks and this will cause data collection to fail.
You could make a copy of the template and change it as you need or create ZBX, if you think this is a mistake.

[rstumbaum]

I did create a haproxy template by myself in 2011 using version 1.8. So yes, I do know my way around and am capable of things like that.

But this is not the point. With every upgrade I try to move previously taken workarounds back into the standard (e.g. used zapcat and it took a while to switch to the jmx agent). Since the standard now provides maintained templates I try again to reduce my own maintenance efforts.

When having a few active-passive haproxies it is inconvenient and failure-prone to set the macro value on a per virtual-ip host entry basis. But you would not assign the agent based template to the underlying machines of the cluster. So from my point of view the argument with the trigger duplication is not valid.

I just wanted to give a feedback to improve the haproxy_http template for everybody.

[eki]

Hi,

is this template works under zabbix version 4.0.3?

Best regards

[interestingusername]

Just in-case anyone trips up like i did;

If you don't perform SSL/TLS termination, the template wont work

[interestingusername]

Okay, on further investigation, it still doesn't work with the majority of stats even when in HTTP mode (like 2xx, 3xx codes)

Also it seems an error to only check 2xx & 3xx for front ends and not back ends (if you host multiple domains, you can only bind :443 to one front end and have to check for different domains)

On even more investigation;

The version template regex is wrong for Ubuntu, you can fix this by changing the template around line 145:

<params>try {
    return value.match(/HAProxy version (\d+\.\d+\.\d+)/)[1];
}
catch (error) {
    throw "HAProxy version is not found : " + error;
}</params>

The code for the JSON path for Nxxx doesn't seem to work, taking this JSON (taken from haproxy csv, coverted to json, shorted for post length);

[
  {
    "pxname": "https-in",
    "svname": "FRONTEND",
    "hrsp_1xx": 0,
    "hrsp_2xx": 263,
    "hrsp_3xx": 406,
    "hrsp_4xx": 2,
    "hrsp_5xx": 16
  }
]

With the following path

$.[?(@.pxname == 'https-in' && @.svname == 'FRONTEND')].hrsp_2xx

it;

Works here: http://jsonpath.herokuapp.com/

Doesn't work here: https://jsonpath.curiousconcept.com/

I cant seem to find anything about which implementation Zabbix use's but it doesn't seem to work ?

Updating the template to

$.[?(@.pxname == '{#PXNAME}')].hrsp_2xx.first()

doesnt fix it either

[max.ch.88]

Okay, on further investigation, it still doesn't work with the majority of stats even when in HTTP mode (like 2xx, 3xx codes)

...

What is the Zabbix, OS and HAProxy version you are using?

[interestingusername]

zabbix 4.4 on ubuntu 18.04.4

HAProxy version 1.8.8-1ubuntu0.9, released 2019/12/02 on Ubuntu 18.04.1

[unficyp]

How can i get the agent based haproxy version to connect to 127.0.0.1:8404 and not to the ip of the server ? i don't want my stats page to be bound to the ip of the host.
thanks

[max.ch.88]

How can i get the agent based haproxy version to connect to 127.0.0.1:8404 and not to the ip of the server ? i don't want my stats page to be bound to the ip of the host.
thanks

You could change {HOST.CONN} to 127.0.0.1 in the item keys.

[max.ch.88]

Hi,

My use case:
I have some haproxy services running on active-passive heartbeat loadbalancers.

When using the haproxy_agent version I get the problem that the inactive host is complaining that the HAProxy service is not running - since it gets started only on failover by heartbeat.

I would suggest using a service like keepalived to have virtual IP and automatic redirection in failover case of active haproxy..

Suggestion:
Could you replace the {$HAPROXY.STATS.HOST} in the item configurations to {HOST.CONN} instead as it is in the haproxy_agent version?

The template has been changed. The link on the file you could find in the header of the topic.

[max.ch.88]

Just in-case anyone trips up like i did;

If you don't perform SSL/TLS termination, the template wont work

Could you show or share your HAProxy configuration so we could make a working template?

[metal]

I am using Zabbix 4.4.8

Any idea why backend(s) session information is not available in the template HAProxy, for example:
Session rate(Current|Max)
Sessions (Current|Max|Total|LbTot|Last)
Bytes (In|Out)



It is essential to have these metrics for backend(s).

These are available for frontend, but not backend, any idea how I can add these items to "(TCP) backend discovery"

thanks