Refactor to use MVCCStore with explicit timestamps #282
Conversation
Summary of ChangesHello @bootjp, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed! This pull request introduces a significant architectural shift in the storage layer by fully embracing a Multi-Version Concurrency Control (MVCC) model with explicit timestamp management. This change standardizes how data is accessed and modified across various service adapters, ensuring that all read and write operations are consistently tied to specific timestamps. The underlying Hybrid Logical Clock (HLC) has been optimized for better concurrency, and the removal of legacy storage implementations streamlines the system towards a unified, MVCC-centric approach. Highlights
Using Gemini Code AssistThe full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips. Invoking Gemini You can request assistance from Gemini at any point by creating a comment using either
Customization To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a Limitations & Feedback Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here. You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension. Footnotes
|
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request is a significant refactoring that moves the system towards a proper MVCC architecture by introducing an MVCCStore with explicit timestamp management. This is a great improvement, simplifying the storage layer and making timestamp handling more explicit. The switch to a lock-free HLC implementation is also a notable performance enhancement. However, I've identified a critical issue in the transaction timestamping logic that could violate atomicity by allowing mixed timestamps within a single transaction. Additionally, there are a couple of high-severity race conditions in the Redis adapter where separate read operations within the same command handler do not use a consistent snapshot timestamp. Addressing these issues is crucial for the correctness and stability of the system.
| if req.IsTxn { | ||
| var startTs uint64 | ||
| for _, r := range req.Requests { | ||
| if r.Ts == 0 { | ||
| if startTs == 0 { | ||
| startTs = i.clock.Next() | ||
| } | ||
| r.Ts = startTs | ||
| } | ||
| } | ||
| return | ||
| } |
There was a problem hiding this comment.
The current logic for assigning timestamps to transactional requests is flawed. If a transaction arrives with a mix of requests—some with timestamps and some without (Ts == 0)—this code will only assign a new timestamp to the latter. This results in a single transaction containing operations with different timestamps, which violates the atomicity guarantees of an MVCC system. All operations within a transaction must share the same start timestamp.
A better approach is to ensure a single, unified timestamp for all requests in a transaction. You could use the timestamp from the first request that has one, or generate a new one if none do, and then apply it to all requests in that transaction.
if req.IsTxn {
var startTs uint64
// All requests in a transaction must have the same timestamp.
// Find a timestamp from the requests, or generate a new one if none exist.
for _, r := range req.Requests {
if r.Ts != 0 {
startTs = r.Ts
break
}
}
if startTs == 0 && len(req.Requests) > 0 {
startTs = i.clock.Next()
}
// Assign the unified timestamp to all requests in the transaction.
for _, r := range req.Requests {
r.Ts = startTs
}
return
}References
- In an MVCC system, all operations within a single transaction must share the same start timestamp to ensure atomicity and a consistent snapshot view.
No description provided.