Bug #54798 data nodes now overly aggressive on memory allocation

• Submitted: 25 Jun 2010 8:27
• Modified: 6 Jul 2010 12:14
• Reporter: Hartmut Holzgraefe
• Status: Won't fix
• Category: MySQL Cluster: Cluster (NDB) storage engine
• Severity: S3 (Non-critical)
• Version: mysql-5.1-telco-7.0
• OS: Linux
• CPU Architecture: Any
• Tags: mysql-cluster-7.0.11 and up

[25 Jun 2010 8:27] Hartmut Holzgraefe

Description:
After the fix for bug #37430 "LockPagesInMainMemory only locks pages after they've been accessed at least once" ndbd does now touch each memory page right after allocation to make sure it is physically present in main memory for locking to take effect.

Problem is that this happens even when LockPagesInMainMemory=0, so putting extra memory stress on the the system on ndbd startup without real need ...

How to repeat:
Check resident set size (RSS) for ndbd right after node start for versions before and after 7.0.11

Suggested fix:
Only touch allocated pages if LockPagesInMainMemory > 0

[6 Jul 2010 12:14] Jørgen Austvik

Won't fix: Part of another fix (bug#37437) that is better.