Use fdatasync (just for testing over an FS)

This commit is contained in:
Vitaliy Filippov
2019-11-27 02:41:30 +03:00
parent 06634054c9
commit ce5cd13bc8
4 changed files with 9 additions and 9 deletions
+3 -3
View File
@@ -7,9 +7,9 @@
//
// This makes 1 4K small write+sync look like:
// 512b+4K (journal) + sync + 512b (journal) + sync + 4K (data) [+ sync?] + 512b (metadata) + sync.
// WA = 2.375. It's not the best, SSD FTL-like redirect-write with defragmentation
// could probably be lower even with defragmentation. But it's fixed and it's still
// better than in Ceph. :) except for HDD-only clusters, because each write results in 3 seeks.
// WA = 2.375. It's not the best, SSD FTL-like redirect-write could probably be lower
// even with defragmentation. But it's fixed and it's still better than in Ceph. :)
// except for HDD-only clusters, because each write results in 3 seeks.
// Stabilize big write:
// 1) Copy metadata from the journal to the metadata device