[
https://issues.apache.org/jira/browse/HBASE-10241?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13857600#comment-13857600
]
Sergey Shelukhin commented on HBASE-10241:
------------------------------------------
bq. sequenceid is not scoped to the region. Ergo mvcc should be too?
That is not necessary, strictly speaking.
bq. IMO, just say no to more configs. When would someone want 'inconsistent
scanners'? (In my experience, users do not play with configs in 98.1745% of
cases – smile).
That is ok, but in 1.8265% cases when you do need config you normally really
need it :)
It can be on by default, but at least for the first version the new
functionality should be easy to disable if needed.
{quote}
bq. ....unless mvcc and seqId are merged)
Would this issue be better if above work is done first?
{quote}
Not much. After examining stuff I'd say it almost doesn't matter. Except maybe
some more work adding mvcc to WAL, but that's not much.
...Recovery will have to make use of mvcc when replaying the edit to new
server.
bq. This work is ongoing over in @jeffrey zhong replay wal effort?
It will have to go into both that and old replay I guess, similar to nonces.
There's also HBASE-10227, which is related.
bq. mvcc is about giving you a consistent view on a row only. This work is to
deal with the case where you have a wide row and you have already passed the
client the first half of a row, a crash happens, and you need to return to the
client the second half of the row?
Actually, mvcc currently gives you a consistent view of the region, in theory
(unless there's some glitch in code). All updates in the region are
mvcc-ordered.
bq. A section in doc. on implications of not having this change fixed and then
the 'cost' of this fix going in would help.
Will add. What do you mean by cost?
> implement mvcc-consistent scanners (across recovery)
> ----------------------------------------------------
>
> Key: HBASE-10241
> URL: https://issues.apache.org/jira/browse/HBASE-10241
> Project: HBase
> Issue Type: New Feature
> Components: HFile, regionserver, Scanners
> Affects Versions: 0.99.0
> Reporter: Sergey Shelukhin
> Assignee: Sergey Shelukhin
> Attachments: Consistent scanners.pdf
>
>
> Scanners currently use mvcc for consistency. However, mvcc is lost on server
> restart, or even a region move. This JIRA is to enable the scanners to
> transfer mvcc (or seqId, or some other number, see HBASE-8763) between
> servers. First, client scanner needs to get and store the readpoint. Second,
> mvcc needs to be preserved in WAL. Third, the mvcc needs to be stored in
> store files per KV and discarded when not needed.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
