Percona Live Online

Your databases and monitoring are all set up and you've got your MySQL, PostgreSQL and MongoDB databases figured out - you're monitoring them and everything is fine. But now you've been tasked to keep tabs on that new Cassandra cluster your company has - we'll show you how to incorporate monitoring it into the Percona Monitoring and Management tool and which features enable you to get the best out of any new and existing database you're incorporating. Database problems? Not on your watch.

Distributed clusters might encounter performance problems or unpredictable failures, especially when they are running in the cloud. Of all the kinds of failures, kernel failures may be the most difficult to analyze and simulate.

Based on Berkeley Packet Filter (BPF), BCC (BPF Compiler Collection) offers many useful resources to construct effective kernel tracing and manipulation programs but it might cause inconveniences for developers in certain situations. Compared with BCC, libbpf + BPF CO-RE seems a better solution because it greatly reduces storage space and runtime overhead, which enables BPF to support more hardware environments, and it optimizes programmers' development experience.

In this talk, Wenbo Zhang will share his BPF practices on how to develop BPF tools with libbpf + BPF CO-RE. He will introduce the advantages of this development method, how to use this method to develop tools, and some tips and tricks for writing Linux BPF applications with libbpf.

Recently, PMM replaced Prometheus with VictoriaMetrics. In the talk we want to cover the motivation behind this transition, the architecture and internals of PMM and technical details of the replacement.
The talk is going to be held by members of both organizations who took a part in migration: Percona and VictoriaMetrics. We expect the talk to be divided into the following parts:
1. The evolution path of PMM and decision to replace Prometheus (by Percona members)
2. PMM architecture and technical details of the transition to VM (by VM members)
3. The summary and results of collaboration

Some key slide titles from the talk:
1. Architecture of PMM
2. Why we decided to replace Prometheus
3. Transition period - what to do with historical data
4. Mutually beneficial collaboration for PMM and VM

The talk will also include the following observability topics:
1. Push vs Pull metrics collection approaches
2. Efficiency of monitoring systems

The popularity of DBaaS cannot be denied. They are incredibly helpful for growth not only due to the operational tasks assistance but also for monitoring and visibility of database internals. In the case of RDS, Amazon provides pretty cool features other than the well-known CloudWatch: things like Enhanced Monitoring or Performance Insights are fantastic....but they came with a (unusually high) cost.

Enter PMM: The Percona Monitoring and Managing tool. PMM being highly customizable and based on well-known open source tools, appears as a great alternative, especially for DBA teams that require deep understanding of what is going on inside the databases.

However, PMM also require a considerable amount of time and effort to have it the way we wanted, especially for PostgreSQL.

Our journey involves work on several aspects like:
- PMM server capacity
- Limitations due to being at a DBaaS
- Several dashboard customization
- Additional data sources via custom queries and textfile-collectors
- Grafana tune
- Prometheus magic
- And some hacking...