EluSiOn wrote:
兩位對 storage 架構真的了解不夠細膩. LVM 在 rebuild 的時候, 都需要call /sbin/fsck.btrfs
為何rebuild raid需要call fsck.btrfs?
換作在ext4上, 難道你把/sbin/fsck.ext4拿掉, raid就無法重建?
建議你先去實驗看看再來下結論
betoptic wrote:
為何rebuild raid...(恕刪)
EluSiOn wrote:
官網的說的很明白, 使用 BTRFS on LVM Synology 這種架構, 會有寫入效能的損失. 官方的開發者說的很明白, 清楚, 這樣子的架構會讓 BTRFS 寫入的優化完全失效. ineffective 是一個超級負面的字 (not producing any significant or desired effect)
EluSiOn wrote:
現在很多企業級儲存系統, 根本不使用 lvm. Oracle 甚至說
we recommend that Logical Volume Manager (LVM) not to be used for the boot disk.
EluSiOn wrote:
The filesystem is not relevant before creating the array. mdadm does not "clobber" both drives, it just copies the contents of the primary to the secondary when mirroring. If the primary already had an ext4 filesystem in it, then it will still look like it does, and after the resync , so will the second. The new default mdadm metadata format of 1.2 is stored 4k from the start, so it will have overwritten and damaged the filesystem that was there before.
如果只讀到前面一半, 後面一半沒有讀! 那很恐怖的!
AristotleC wrote:
不過實際上就是 Btrfs RAID 很不爭氣的,要嘛不能用,不然就是效能測出來跟配 RAID/LVM 不相上下,看來離這些開發者的夢想還有一段距離。
AristotleC wrote:
另外 LVM 其實很常被使用,在 Redhat 的維護下也一直很穩定,要戰 Btrfs 應該也不用把 LVM 拖下水啊。
EluSiOn wrote:
你的說法跟 phoronix...(恕刪)