Are you working with the NoSQL storage Apache Cassandra?
On May 23, Odnoklassniki invites experienced developers to their office in St. Petersburg for , dedicated to working with Apache Cassandra. Only your experience with Cassandra and your willingness to share it are important.
We at OK Apache Cassandra in 2010 to store photo ratings. Currently, we are the largest users of Apache Cassandra in the RuNet and among the largest in Europe. We have over a hundred different clusters used for storing various product information - classes, chats, messages, as well as managing critical infrastructure data - mapping logical blocks to disks of large binary storage - , managing internal cloud data. etc.
In total, under the management of Cassandra, there are petabytes of data across thousands of nodes. Over this time, we have accumulated vast experience in administration, development, and operation of solutions based on Cassandra and even developed our own .
Now we would like to share all of this with you - through real cases from practice and without secrets; The event will take the form of a live discussion among participants, which means that the discussion will take up most of the time. are ready to share their ideas and approaches. The event will be led by and .
What topics will be covered?
Operation:
We will review typical configurations of nodes and clusters in various production installations. We will discuss how to scale clusters as data volume and load increase, and how to replace failed nodes with minimal impact on clients. We will share pain points and systematize popular pitfalls. We will find out how to monitor clusters to understand ahead of time where and what exactly is not working. We will touch on the problems of deploying new versions of Cassandra.
Performance:
We will try to understand which metrics to look at and what can be fine-tuned to improve metrics. We will discuss whether to retry or not and if so, how to do it. We will identify bottlenecks in the architecture and implementation of Cassandra and review some engineering tricks to bypass them. We will touch on the pressing issue of regular repair and compaction without degrading performance.
Fault tolerance:
Hardware isn't everlasting, so failures happen frequently, and a colleague's hand may tremble, leading us to delete something unnecessary. Let's discuss recovery from disk, machine, or data center failures, as well as restoring to a consistent state from backups in cases of operator errors.
And tell your friends and colleagues about the event.
Source: habr.com
