AG stride testing was performed on a system with ample amounts
of memory, so prefetching with AG stride during Phase 4 was
missed. The attached patch fixes this.
32 AGs, ag_stride = 4:
Phase 3 - for each AG...
- scan and clear agi unlinked lists...
- process known inodes and perform inode discovery...
- agno = 0
- agno = 4
- agno = 8
- agno = 12
- agno = 16
- agno = 20
- agno = 24
- agno = 28
which is correct... but in Phase 4:
Phase 4 - check for duplicate blocks...
- setting up duplicate extent list...
- check for inodes claiming duplicate blocks...
- agno = 0
- agno = 1
- agno = 2
- agno = 5
- agno = 6
- agno = 4
- agno = 7
- agno = 3
fix_ag_stride
Description: Binary data
|