[Sheepdog] [PATCH 0/4] some optimization for cluster recovery.
levin108 at gmail.com
Fri Mar 30 05:12:41 CEST 2012
In this patch set, I did the following work:
1. By only writing objects no longer belong to the current node
to the snapshot, avoid unnecessary IO operation in farm when cluster
2. Add a nr_zones field to cluster_info, which is maintained by sheep,
so we can use it anywhere instantly, instead of recalculating it.
3. Add a cleanup operation for collie for users to clean up the useless
trunk files in farm.
4. No longer store the trunk map in file, instead, I put it in memory
using rb-tree, after restart, we just need to clean up the old trunk
files by the cleanup function implemented in patch 3.
More information about the sheepdog