[
https://issues.apache.org/jira/browse/HBASE-50?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12876352#action_12876352
]
stack commented on HBASE-50:
----------------------------
On, "5 Snapshot Creation"
.bq "Because this table region must be online, dumping the HRegionInfo of the
region to a file ".regioninfo" under the snapshot directory of this region will
obtain the metadata."
...The above is wrong, right? We can snapshot online tables?
+1 on reading .META. data, flushing it to .regioninfo to be sure you have
latest, and then copying that (Or, instead, you could ensure that on any
transistion, the .regioninfo is updated. If this is happening, no need to do
extra flush of .META. at snapshot time. This latter would be better IMO).
So, do you foresee your restore-from-snapshot running split over the logs as
part of the restore? That makes sense to me.
Why you think we need a Reference to the hfile? Why not just a file that lists
the names of all the hfiles? We don't need to execute the snapshot, do we?
Restoring from a snapshot would be a bunch of file renames and wal splitting?
Or what are you thinking? (Oh, maybe I'll find out when I read chapter 6).
.bq ....can be created just by the master.
Lets not have the master run the snapshot... let the client run it?
Shall we name the new .META. column family snapshot rather than reference?
I like this idea of keeping region snapshot and reference counting beside the
region up in .META.
On the filename '.deleted', I think it a mistake to give it a '.' prefix
especially given its in the snapshot dir (the snapshot dir probably needs to be
prefixed with a character illegal in tablenames such as a '.' so its not taken
for a table directory).
Regards 'Not sure whether there will be a name collision under this ".deleted"
directory', j-d has done work to ensure WALs are uniquely named. Storefiles
are given a random-id. We should probably do the extra work to ensure they are
for sure unique... give them a UUID or something to we don't ever clash.
After reading chapter 6, I fail to see why we should keep References to files.
Maybe I'm missing something.
.bq Not decides where to keep all the snapshots information, in a meta file
under snapshot directory
Do you need a new catalog table called snapshots to keep list of snapshots, of
what a snapshot comprises and some other metadata such as when it was made,
whether it succeeded, who did it and why?
On the other hand, a directory in hdfs of files per snapshot will be more
robust.
Section 7.4 is missing split of WAL files. Perhaps this can be done in a MR
job?
Design looks excellent Li.
> Snapshot of table
> -----------------
>
> Key: HBASE-50
> URL: https://issues.apache.org/jira/browse/HBASE-50
> Project: HBase
> Issue Type: New Feature
> Reporter: Billy Pearson
> Assignee: Li Chongxin
> Priority: Minor
> Attachments: HBase Snapshot Design Report V2.pdf, snapshot-src.zip
>
>
> Havening an option to take a snapshot of a table would be vary useful in
> production.
> What I would like to see this option do is do a merge of all the data into
> one or more files stored in the same folder on the dfs. This way we could
> save data in case of a software bug in hadoop or user code.
> The other advantage would be to be able to export a table to multi locations.
> Say I had a read_only table that must be online. I could take a snapshot of
> it when needed and export it to a separate data center and have it loaded
> there and then i would have it online at multi data centers for load
> balancing and failover.
> I understand that hadoop takes the need out of havening backup to protect
> from failed servers, but this does not protect use from software bugs that
> might delete or alter data in ways we did not plan. We should have a way we
> can roll back a dataset.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
