Cluster Snapshot Consistency via Change Tracker and Merge
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems face challenges in creating consistent cluster-wide snapshots across multiple storage sites in a virtualized environment, often resulting in delays and inconsistencies in snapshot timeliness.
Innovation Solution
A method is implemented to track input/output (I/O) metadata across multiple sites, where point-in-time copies are created separately at each site, and then merged at a target site to ensure consistency, involving change trackers that stop tracking after snapshot creation and are later restarted.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If point-in-time copies are created separately at each site, then snapshot creation speed is improved, but snapshot consistency across sites deteriorates
Solution Approach 1:
The system performs preliminary actions by tracking I/O metadata before snapshot creation, recording all changes that occur at each site. This advance tracking enables the subsequent merging process to efficiently reconcile differences between sites without requiring simultaneous snapshot creation, thus maintaining both speed and consistency.
Solution Approach 2:
The patent introduces an intermediary merging process that receives point-in-time copies from multiple sites and reconciles them into a consistent cluster-wide snapshot. This intermediary step acts as a mediator that resolves inconsistencies between separately created snapshots while minimizing overall processing time.
2Adaptability or versatility
If change trackers are restarted after snapshot creation, then I/O tracking resumes for future snapshots, but timing and coordination complexity increases
Solution Approach 1:
The system discards the change tracker state after it has served its purpose of tracking I/O metadata for snapshot creation. After the snapshot is created and merged, the tracker is recovered or restarted to begin tracking again for future snapshot operations. This cyclic discard and recover process maintains simplicity while enabling continuous tracking capability.
Data Source
AI summary
The described system provides that backend array-based snapshots may be created separately on each site of a cluster, and then the snapshots fixed so as to be consistent and/or otherwise identical among the plurality of sites. The system advantageously allows creation of a consistent cluster-wide snapshot with minimal or no I/O delays. In an embodiment, the system provides for use of a change tracker that tracks the metadata of all the I/Os incoming to the volumes being snapped. When the system wants to create a snapshot on all sites, the change tracker is activated on each site separately for the volumes being snapped. A snapshot is then created on each of the cluster sites/backend storage arrays separately. The change trackers are then ordered to stop tracking. A snapshot fixing procedure is then initiated.