Optimize undo spaces lock

[wjunLu]

Analysis

During undo tablespace truncation in MySQL, when the target undo tablespace has grown significantly large, the undo_spaces_lock (a read-write lock) experiences severe contention, as demonstrated below

The undo_spaces_lock safeguards concurrent access to the following data structures:

• undo::spaces::m_spaces
• undo::space_id_bank
• undo truncation logs

Performance Impact:
When the purge coordinator thread holds the undo_spaces_lock in write mode, all DML foreground threads become blocked, resulting in measurable throughput degradation during this period.

Critical Section Analysis:
The purge coordinator thread pool acquires the lock solely to protect:

• undo::space_id_bank
• undo truncation logs
  (Note: It never modifies undo::spaces::m_spaces)

DML foreground threads require the lock only to ensure: read consistency of undo::spaces::m_spaces
(Key Insight: No inherent logical conflict exists between these thread types)

Optimization Strategy

1. Snapshot Mechanism:

• After obtaining the write lock, the purge coordinator creates a snapshot of undo tablespaces
• DML threads then query this snapshot directly, eliminating their need for read locks

2. Fine-grained Synchronization:

• Replaces coarse-grained locking with atomic operations for:
  • Version tracking
  • State synchronization
• Maintains consistency between purge coordinator and DML threads

Figure 1 - current undo_spaces_lock (Before)

Figure 2 - The optimized undo_spaces_lock (After)

[mysql-oca-bot]

Hi, thank you for submitting this pull request. In order to consider your code we need you to sign the Oracle Contribution Agreement (OCA). Please review the details and follow the instructions at https://oca.opensource.oracle.com/
Please make sure to include your MySQL bug system user (email) in the returned form.
Thanks