snapshot delete doesn't free space
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
sheepdog |
Fix Committed
|
High
|
Hitoshi Mitake |
Bug Description
Description:
each time we take a snapshot we "freeze" some data so the cluster used space grows.
Snapshot can be deleted successfully but the relative space isn't released.
Ho to reproduce:
E.g. a guest using a vdi of 10G.
If this is the only vdi in the cluster, 'Total virtual image size' will be 10G and the total used space will be not more than 20G.
(From 'dog node info').
start the guest
dog vdi snapshot vdiName
write data inside the guest almost till the end of the available space
dog vdi snapshot vdiName
remove the data you have been writing
dog vdi snapshot vdiName
repeat these steps as much as you wish.
Run 'dog node info'.
Total used space may be something like 35G.
Removing all snapshots, it should go back to 20G, but it doesn't.
Consequences:
depending on how often you take snapshot and how much data you write on the guest side, you may fill up your cluster.
Considerations:
this issue is know since long time and should be fixed by 'object reclaim patch'.
Changed in sheepdog-project: | |
status: | New → In Progress |
Changed in sheepdog-project: | |
assignee: | nobody → Hitoshi Mitake (mitake-hitoshi) |
Changed in sheepdog-project: | |
importance: | Undecided → High |
Changed in sheepdog-project: | |
status: | In Progress → Fix Committed |