The article analyzes Pi 1.0 and Pi Durable, a minimalist agent harness framework that now supports MCP via Codemode, introduces virtual model routing, and provides six durability primitives including crash recovery, session forking, and background compression for long-running, multi-user agent applications.
This article compares Pi and DeepSeek Harness agent loop architectures across four control points—input attribution, tool scheduling, closure permissions, and persistent evidence—revealing how each handles mid-task user constraints, parallel tool execution, token truncation, and crash recovery through distinct runtime contracts and durable extensions.
This article explains InnoDB's checkpoint mechanism, contrasting synchronous and fuzzy checkpoints, detailing the pre-MySQL 8.0 fuzzy checkpoint implementation using flush lists and page cleaners, and describing two principles for detecting abnormal shutdowns plus four post-restart crash scenarios with their recovery requirements.
ElephpantDB is an experimental SQL database implemented in a single PHP file, supporting CREATE TABLE, SELECT, UPDATE, DELETE, indexes, prepared statements, an append-only heap with crash recovery, hash indexes, and an optional HTTP JSON endpoint, running on PHP 8.3+.
PI's 3,446-line harness-v2.md redesigns the agent runtime around crash recovery as the primary constraint, using intent records with provisioned IDs, a four-layer session state, persisted retry counts, deferred writes to preserve provider KV cache, and exhaustive crash-point testing via manual drive mode.
The article explains the design of Hermes' terminal UI gateway, detailing why the TUI does not import the Agent directly, how JSON‑RPC over stdio or WebSocket enables local and remote modes, and how slash commands are processed through a two‑layer routing system with a dedicated _SlashWorker subprocess and a pure‑function crash‑recovery plan.
The article explains that MySQL redo log operates at the InnoDB engine layer to ensure transaction durability and crash recovery, while binlog works at the server layer to record logical changes for replication, archiving, and point‑in‑time recovery, highlighting their distinct layers, purposes, content, and write mechanisms.
This article provides detailed explanations of ten core MySQL topics—including InnoDB lock upgrades, redo log mechanics, MVCC version chains, tablespace types, page splitting, optimizer plan selection, crash recovery, binlog vs redo log, adaptive hash index, and memory management—along with practical tips for avoiding pitfalls and tuning performance.
The article analyzes a MySQL 5.7 replication bug where, after consecutive crashes, the slave ends up with more GTIDs than the master because the last binlog’s GTIDs are not persisted, and demonstrates the problem using GDB debugging and step‑by‑step reproduction.
This article explains MySQL's binary log (binlog), a binary format file that records SQL statements for database updates, its role in master-slave replication, crash recovery, and detailed implementation mechanisms including cache management and event writing processes.
This article explains how InnoDB uses undo segment states for crash‑safe recovery, details the XA transaction protocol, describes bugs in earlier MySQL versions, and shows how MySQL 8.0.30 introduces a new undo state and revised prepare order to make external XA transactions crash‑safe, including code examples and recovery steps.
This article explains the design of Virtio-fs, its architecture and high‑availability features, and details the crash‑recovery mechanism—including crash models, state preservation, supervisor coordination, request idempotence, downtime optimization, and hot upgrade/migration—implemented by ByteDance's STE team for secure container workloads.
This article explains MySQL's crash recovery process, covering how the doublewrite buffer repairs corrupted pages, how redo logs are applied to restore data, and how undo tablespaces and unfinished transactions are handled during server startup.
This article explains how MySQL performs crash recovery by using Redo logs, doublewrite buffers, undo tablespace handling, transaction subsystem initialization, and the processing of active, prepared, and committed transactions to restore data pages and ensure consistency after an unexpected shutdown.
This article explains MySQL's layered architecture, the roles and mechanisms of redo log and binlog, their write and flush strategies, crash‑safe guarantees, and how two‑phase commit ensures consistency, providing practical commands and configuration tips for reliable database operation.
This article explains MySQL's two essential logs—redo log and binlog—by describing the database's layered architecture, the write‑ahead logging mechanism, persistence strategies, crash‑safe guarantees, and how a two‑phase commit ensures consistency between them.
This article explains how PolarDB ensures transaction atomicity by analyzing ACID principles, detailing the shortcomings of simple Direct‑IO and Buffer‑IO, introducing write‑ahead logging, full‑page backup, checkpointing, visibility mechanisms, and distinguishing durability and atomicity points to achieve reliable crash recovery and consistent data visibility.
This article explains the ZAB (ZooKeeper Atomic Broadcast) protocol, detailing its atomic broadcast mechanism, two‑phase commit style messaging, transaction IDs (ZXID), leader‑follower coordination, crash recovery principles, data synchronization, and the underlying ZXID design with illustrative code snippets.
This article investigates why MySQL 5.7 crash recovery can take hours on a development instance with thousands of tables, detailing stack‑frame analysis, GDB debugging, InnoDB source inspection of validation logic, performance testing, and methods to bypass the table‑space verification step.
This article explains the ZAB protocol—Zookeeper's atomic broadcast and crash‑recovery mechanism—detailing its design, message‑broadcast process, leader election, transaction ordering with ZXID, and how it ensures data consistency and availability in distributed systems.
This article explains Oracle instance recovery, detailing when it is needed, how checkpoints and SCNs are recorded during clean and abnormal shutdowns, and the step‑by‑step forward‑roll and rollback process performed by SMON to restore database consistency after a crash.
