As did for the last year edition, our DevOps team participated at InfluxDays, a conference organized by InfluxData and focused on time series data. We were really looking at the event with excitement, for all the new features InfluxData is putting into the new InfluxDB 2.0. So, apart from grabbing a very nice t-shirt, we attended a lot of interesting talks.
Gianluca Arbezzano from InfluxData spoke about how to monitor a Kubernetes based infrastructure using InfluxData technology. In the case of Kubernetes, we need to be sure to monitor not only the hardware on which all the K8s infrastructure is running on, but also any service, ingress, and pod that has been deployed. To solve the first issue, Telegraf can be deployed as DeamonSet to gain visibility on every host that compose our K8s cluster. The Telegraf agent monitors resources like iops, network, cpu, memory, disk and other services from the host. To solve the second issue, we can resort again to Telegraf, but this time in sidecar mode. Telegraf can then installed as sidecar in the pods to monitor, guaranteeing that our monitoring configuration is very close to each application. At this point, all your container and cluster data can be stored in InfluxDB and manipulated via Flux with almost zero pain.
Luca Deri from NTOP gave us a cool insight into network traffic monitoring in the era of containers. Traditionally, tools analyze each incoming and outgoing packet. In an architecture based on containers, packet analysis like was in the past is no longer sufficient to have visibility on how services can interact inside a system. Luca explained to us how open source tools developed by NTOP can provide the capability of monitoring containers. This can be done with the help of InfluxDB: events and network interaction between services and containers retrieved by the various NTOP tools are stored as time series into InfluxDB.