Zabbix and Containerized Environments

[Yello]

Hi,
I'm wondering how well zabbix plays in containerized environments? I ask because the focus of our monitoring is more on application and service level heuristics and not just infrastructure level metrics. Therefore, how an instance is monitored can vary quite significantly depending on the service. Plus, we also have the need to allow developers to define and deploy their own checks...

It strikes me that the zabbix configuration model may not be nimble enough to keep pace with a fast changing cloud infrastructure or to support developers defining their own configuration in an easy to use way. Am I conceiving a use case that goes beyond what zabbix is targeted at? I'm not sure which is why I'm asking.

I like zabbix mainly because of the level of control it gives over alerting. Have others been able to make efficient use of zabbix in the kind of changeable, federated environment I've described without obstructing devs and their deployments?


Regards,
Dave

[kloczek]

Any type containerization has nothing to do with running software inside those containers.
Zabbix is monitoring software and it can monitor even bunch of dwarfs working in salt mine as long as you can only present parameters of those working dwarfs in forms of numbers or strings.

Generally your dilemmas and questions have nothing to do with forum subject, containers, clouds, support developers defining their own configuration, federated environment (whatever you understand under this term) as well
Your dilemmas and questions are more related to what do you want exactly to monitor .. and our knowledge about those things.

[Yello]

Your dilemmas and questions are more related to what do you want exactly to monitor .. and our knowledge about those things.

I've used zabbix long enough to know that it can monitor anything. That's not up for debate and isn't my question anyway. I'm querying the real world tractability of provisioning zabbix in cloud environments while also trying to give developers control of their own collection. Having them define their own templates, items and alerts in a timely and reliable fashion without impacting deployments and/or having the process of monitoring/dashboard provisioning getting in their way seems a trickier problem than it might first appear.

So while I really like zabbix I'm leaning towards looking at tools like prometheus et.al. to track developer apps and services and leave zabbix to focus on enterprise type monitoring. But, I'm asking the question cos maybe others have done this already or I'm over-emphasizing the points of friction...


Regards,
Dave

[kloczek]

I'm querying the real world tractability of provisioning zabbix in cloud environments while also trying to give developers control of their own collection.

You really need to say what you understand under this tractability.
If your container is running on IP A and what you need is connect from proxy to this IP in case using passive agent or connect from container to serve/proxy. From this point this that exact agent is running in "cloud" is only coincidence. From network communication perspective your system can even run on top of real cloud (somewhere above the ground) and even if network packets will be delivered by pigeons (https://www.ietf.org/rfc/rfc1149.txt https://en.wikipedia.org/wiki/IP_over_Avian_Carriers) still interaction between zabbix agent and server/proxy will be the same as in case non-cloud (whatever you want to understand under such term).

Having them define their own templates, items and alerts in a timely and reliable fashion without impacting deployments and/or having the process of monitoring/dashboard provisioning getting in their way seems a trickier problem than it might first appear.

If inside your container will be running (let's say) java with some application monitoring of such JVM and application will be exactly the same as in case non-cloud env .. because it will be even the same binary which could be used on bare-metal installation.
Really have no idea why you want to use some special/custom templates (?)

So while I really like zabbix I'm leaning towards looking at tools like prometheus et.al. to track developer apps and services and leave zabbix to focus on enterprise type monitoring. But, I'm asking the question cos maybe others have done this already or I'm over-emphasizing the points of friction...

if your developers application will expose somehow version of the app all what you need is read this version and store it in some metric. That's the end of the story ..
Track changes of this exact metrics could provide zabbix in form of raising alarm over trigger or storing md5 checksum and store it as number and observe changes of the value of the metric on the graph.

[Yello]

i think I might see if we can contact zabbix directly and bring them in for a chat. You seem to be looking to debate, in an aggressive and at times condescending tone I might add, completely different issues to those raised...