Snapshot checksum APIs #35
Conversation
Added a new function which generate CRC from ISO standard in reflected format. Longhorn 8666 Signed-off-by: Damiano Cipriani <[email protected]>
A new flag has been added to blob metadata to enable the addition of new xattrs to a read only blob. Longhorn 8666 Signed-off-by: Damiano Cipriani <[email protected]>
Added a new function to create a snapshot with options. Longhorn 8666 Signed-off-by: Damiano Cipriani <[email protected]>
Added a new function to compute sanpshot's data checksum and to store it as an xattr. Longhorn 8666 Signed-off-by: Damiano Cipriani <[email protected]>
Added a new function to create a snapshot with provided options, like a list of xattrs to be set in the snapshot and an option to enable the addition of new xattrs to the snapshot after its creation. Longhorn 8666 Signed-off-by: Damiano Cipriani <[email protected]>
Added an option in snapshot create to enable the addition of new xattrs to a snapshot after its creation. The status of the option is shown in dump_info_json callback. Longhorn 8666 Signed-off-by: Damiano Cipriani <[email protected]>
Added functions to register and get the checksum of snapshot's data. Longhorn 8666 Signed-off-by: Damiano Cipriani <[email protected]>
Added the RPC interfaces to register and to get snapshot's data checksum. Also a lvol test have been added to check the behavior of the new RPCs. Longhorn 8666 Signed-off-by: Damiano Cipriani <[email protected]>
|
Need to update go-spdk-helper as well. |
There was a problem hiding this comment.
I have one question:
After deleting a parent snapshot, IIUC, its clusters should be squashed into the child snapshot. Then, do we need to manually re-calculate the child snapshot checksum?
Also, is this API a sync call? If YES, I am worried that calculating a huge snapshot will block the caller for a long time.
Yes, but if the child snapshot has already a checksum stored, actually we would get an error. What we could do is to add a force option to overwrite the old checksum.
Yes, it is a sync call. We could make it async, and to test the end of the operation we can use the RPC to retrieve the checksum: until the calculation is not ended, we would get an error of xattr not present. |
Which code guarantees this kind of validity of the snapshot checksum?
Do you mean that the caller can open a separate thread to call this API then retrieve if the checksum presents or not? |
Actually no code guarantee the validity of snapshot's checksum. What I mean is that if a snapshot has already the snapshot checksum xattr and we try to calculate again the checksum we will get an error.
Exactly |
Then we need a way to handle this case. Otherwise, this feature may be meaningless, as we cannot prevent users from deleting snapshots hence we never know the snapshot checksum recorded in xattrs is still valid or not. |
A possible solution. |
Will you implement it in this PR? |
Yes, it will the first thing to do after the HackWeek. |
Which issue(s) this PR fixes:
Issue #8666
What this PR does / why we need it:
Special notes for your reviewer:
Additional documentation or context