August 2019 Summaries

The blog post announces the winners of the ScyllaDB University Challenge, highlighting the positive feedback from participants about their experiences with ScyllaDB and its educational platform, ScyllaDB University. Participants praised ScyllaDB for its performance, cost-efficiency, ease of use, and minimal administration needs, with several highlighting the quality and accessibility of the university's online courses. Testimonials from winners such as Dinnemeeda Vishnu Vardhan Reddy, Bhaskar Pandey, Felipe Móz, Rahul S. Gaikwad, and James Caruana commend ScyllaDB for its capabilities in the NoSQL database category, and the initiative's role in spreading knowledge about ScyllaDB. As a token of appreciation, participants received ScyllaDB T-shirts, and the post encourages further engagement through additional courses and community interaction on Slack.

Alexys Jacob, CTO of Numberly, is set to present at the ScyllaDB Summit 2019, where he will discuss the company's practical experiences with MongoDB and ScyllaDB from both development and operations perspectives. Known for his expertise in Big Data architecture and as a strong proponent of open source, Jacob has previously shared insights on Numberly's transition from MongoDB to ScyllaDB, highlighting the use of ScyllaDB for latency and scale-sensitive applications. Numberly continues to use MongoDB for its schema flexibility but has integrated ScyllaDB into various web backends due to its scalable performance. Jacob emphasizes the importance of bridging web, DevOps, and data teams through a common language, notably Python, to achieve cohesive strategies, and he notes recent tech stack updates, including a move to Kubernetes and GraphQL APIs supported by ScyllaDB. Additionally, he will discuss Numberly's approach to change management and staff training in adopting ScyllaDB, aiming to provide actionable insights for both infrastructure and development-focused attendees at the summit.

ScyllaDB Monitoring Stack 3.0 introduces significant updates aimed at improving user experience and efficiency, including a reorganization of dashboards, an upgrade to Grafana version 6, and a cleanup of metrics. The dashboards have been simplified to make navigation easier, with updated names and URLs that better represent their content. The Overview dashboard provides a quick glance at cluster activity and node status, while the Detail dashboard offers in-depth insights into node operations, organized into sections such as Read and Writes, Memory, Cache, and Materialized Views. The update also involves a shift to node_exporter 0.17 for OS metrics reporting, requiring a check for compatibility and potential upgrades. Additionally, the stack now utilizes Python 3 for dashboard generation and genconfig utilities. The updated stack is open-source and available for download on GitHub, with feedback encouraged via private contact or their Slack channel.

ScyllaDB Monitoring Stack 3.0 has been released, offering an open-source solution for monitoring ScyllaDB Enterprise and Open Source versions, using Prometheus and Grafana as its foundation. This new release supports multiple versions of ScyllaDB and introduces several enhancements, including a general reorganization of dashboards, updated metrics and labels, and the removal of certain configuration files to streamline deployment customization. Notably, the update involves a switch from Grafana 5 to Grafana 6, providing access to newer features, and from Python 2 to Python 3 due to the former's impending end of life. Additional improvements include an upgrade from Prometheus 2.7.2 to 2.10, the introduction of new default alerts for disk space and node status changes, and real-time display of each node’s version in the Nodes table. Users are advised to follow the upgrade guide closely due to potential backward compatibility issues and changes in node_exporter requirements.

ScyllaDB Open Source 2.3.6 is a bugfix release of the 2.3 stable branch, announced by the ScyllaDB team, and it maintains backward compatibility with support for rolling upgrades. Users are encouraged to upgrade to the latest stable branch, ScyllaDB 3.0, for enhanced features. The release addresses several stability issues, including potential system hangs, segmentation faults during system size estimates queries, and undefined behavior in sstables cleanup. A new command-line flag, --abort-on-internal-error, has been introduced to facilitate error detection and debugging by exiting the system promptly when such errors occur. Additionally, the update resolves issues with the view builder process and performance impacts in specific scenarios of the Time Window Compaction Strategy.

ScyllaDB Open Source 3.0.10 is a bugfix release from the ScyllaDB 3.0 stable branch, ensuring backward compatibility and supporting rolling upgrades. This version addresses several stability issues, including race conditions between compaction and sstable reshuffling, system hangs or segmentation faults during system size estimation queries, and problems with command-line option parsing for Docker. Other fixes involve handling unclosed partition format errors, aborting on internal errors, and addressing failures during streaming and node restarts. The release also resolves undefined behaviors in sstables cleanup, issues with MC sstable format, potential deadlocks in connection authentication, and exceptions in index readers, enhancing the overall stability and reliability of the system.

The webinar hosted by ScyllaDB Field Engineer Juliana Oliveira offers insights into best practices for data modeling specifically tailored for ScyllaDB, emphasizing the importance of understanding data storage and distribution within the database. Oliveira contrasts SQL with ScyllaDB's Cassandra Query Language (CQL), highlighting the differences in data modeling approaches, where ScyllaDB relies on denormalization and data organization based on query patterns. The session delves into partitioning and clustering keys, using a veterinary clinic example to demonstrate how data is sorted and queried, and addresses potential issues like large and hot partitions. Furthermore, it explores ScyllaDB’s underlying storage mechanics, including the use of memtables and SSTables, and discusses various compaction strategies such as size-tiered and leveled compaction, which impact performance and resource utilization. The webinar aims to equip attendees with foundational knowledge to optimize data distribution and manage partition sizes effectively, enhancing database performance and efficiency.

ScyllaDB Enterprise 2019.1.2 is a patch release focused on enhancing the stability and robustness of the ScyllaDB Enterprise 2019.1 branch by addressing various issues and adding mechanisms for better error handling and reporting. This release includes fixes for potential race conditions, unhandled exceptions, undefined behaviors, and segmentation faults, as well as improvements in the handling of streaming, SSTable operations, and connection authentication. It also addresses specific CQL-related errors and performance issues, such as redundant RPC connections during repairs and inefficiencies in large collection operations. The update encourages ScyllaDB Enterprise customers to coordinate with the support team for upgrading to this version, which is designed to maintain the high performance required for real-time big data workloads.

ScyllaDB Enterprise 2018.1.14, released by the ScyllaDB team, is a production-ready patch update for the stable 2018.1 branch, aimed at fixing a specific issue where the database might exit unexpectedly when using the Leveled Compaction Strategy to flush memtables. This release serves as a bug fix for an issue that was introduced in version 2018.1.12 and is particularly designed for handling real-time big data workloads in ScyllaDB's high-performance enterprise NoSQL database framework. ScyllaDB Enterprise customers are advised to upgrade to this version with the assistance of the ScyllaDB support team, and while 2018.1.14 addresses this immediate concern, users are also encouraged to consider upgrading to the latest version, 2019.1, for more comprehensive updates.

ScyllaDB Open Source 3.1 introduces a significant improvement in its data synchronization process with the implementation of a row-level repair algorithm, which enhances performance by operating at a more granular level compared to the previous partition-level approach. This new method minimizes data transfer by exchanging only mismatched rows between nodes, rather than entire partitions, thereby reducing unnecessary data transmission and resource consumption. Additionally, the new repair process is more efficient, achieving faster synchronization times and reduced data transfer costs, particularly in scenarios where only a small percentage of data is out of sync. Benchmark results demonstrate considerable improvements, with the new algorithm being up to 6.78 times faster in some cases and using significantly less bandwidth. The enhancements also include increased parallelism and the use of a faster, non-cryptographic hash, resulting in reduced CPU, disk, and network resource usage and freeing up resources for online requests. These changes not only improve the efficiency and speed of repairs but also reduce the likelihood of failures during the repair process, with further improvements planned for future ScyllaDB versions.

At the ScyllaDB Summit 2018, Google’s Holden Karau presented on building resilient pipelines in Apache Spark, addressing the challenges of Spark failures and outlining strategies for recovery. As an open-source advocate and Spark committer, Karau highlighted the complexities and redundancies within Spark’s components, such as multiple machine learning and streaming engines, which can complicate pipeline recovery. She demonstrated the construction of a recoverable Wordcount example to illustrate the process, emphasizing the importance of checkpoints and handling success markers to mitigate failure impacts. Despite Spark's lazy evaluation leading to late error detection, Karau proposed techniques like caching and asynchronous saving to improve resilience, while also acknowledging the limitations and potential inefficiencies of these solutions. She stressed the importance of using job IDs to manage separate backfills without data interference and concluded by underscoring the necessity of testing and focusing recovery efforts on critical parts of the pipeline.

Apache Spark, a unified analytics engine for large-scale data processing, was a key topic at the ScyllaDB Summit 2018, where Eyal Gutkind from ScyllaDB shared insights on best practices for integrating Spark with ScyllaDB in heterogeneous data environments. Gutkind emphasized the importance of understanding the interplay between cluster configurations to enhance resilience in long-running analytics jobs and addressed common challenges such as deployment in diverse big data ecosystems and optimizing analytics workloads. He provided strategies for efficiently deploying Spark with ScyllaDB, such as properly sizing nodes, tuning partition sizes, and configuring concurrency settings and retry policies to improve data processing and reduce latency. Gutkind also discussed the architectural differences between Spark and ScyllaDB, particularly in data sharding and batch processing, highlighting that Spark's lazy data consumption contrasts with ScyllaDB's even data distribution across nodes. The session underscored the importance of resource management and system tuning to achieve optimal performance, suggesting that separating Spark and ScyllaDB clusters could lead to improved efficiency and performance.

ScyllaDB has released two new versions, 2018.1.12 and 2018.1.13, of its ScyllaDB Enterprise software, focusing on bug fixes to enhance stability and performance. Version 2018.1.12 addresses issues with the Gossip protocol, improving stability when adding nodes to a cluster, while 2018.1.13 fixes a crash occurring in rare cases with the old SSTable file format. Both updates include various other fixes, such as resolving memory allocation issues, schema change delays, and performance regressions. Users are encouraged to upgrade to these new versions in collaboration with the ScyllaDB support team to ensure optimal performance and reliability.

ScyllaDB Open Source 3.0.9 has been released as a bugfix update for the stable 3.0 branch, maintaining backward compatibility and supporting rolling upgrades. This release addresses several issues, including redundant RPC connections during shard range repairs, marshalling errors with timezone formats in CQL, and potential races between role_manager and password_authenticator, which can cause ScyllaDB to exit. Additionally, it resolves problems with using tuples as clustering keys without to_string() implementations and fixes segmentation faults when replacing expired SSTables. Users are encouraged to report any encountered issues, and various installation options are available, including Docker, binary packages, and EC2 AMI.