Implement "awaiting recovery" mode for redundant ZEORaid servers
Bug #485210 reported by
Christian Theune
This bug affects 1 person
| Affects | Status | Importance | Assigned to | Milestone | |
|---|---|---|---|---|---|
| gocept.zeoraid |
Confirmed
|
Wishlist
|
Unassigned | ||
Bug Description
As reported in #484727 we currently have to disable redundant ZEORaid servers while recovering a storage.
IMHO we could allow keeping the redundant ZEORaid servers running while recovering a storage if we do something smart like this:
- Signal N-1 ZEORaid servers that we are about to recover a storage. This will cause those ZEORaid servers to open the storage we're about to recover.
- Signal the 1 remaining ZEORaid server to recover that storage.
- While the N-1 ZEORaid servers are waiting, periodically (e.g. at transaction boundaries or something similar) check whether the recovered storage is good again and perform (part of or a similar job to) the finalize recovery procedure.
| Changed in gocept.zeoraid: | |
| importance: | Undecided → Wishlist |
| Changed in gocept.zeoraid: | |
| status: | New → Confirmed |
To post a comment you must log in.
