[
https://issues.apache.org/jira/browse/HBASE-2856?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
stack updated HBASE-2856:
-------------------------
Assignee: stack (was: ryan rawson)
Ryan, Kannan, and myself chatted on this issue a while:
+ We'll add a new insertion sequence id into the KeyValue Key. It will be
inserted after the current Timestamp and before KeyValue Type. This new
insertion sequence id will be used resolving 'order' when all else matches in
two KeyValue keys.
+ We'd use the last two bits of the KeyValue Type for 'version'. We'll set bit
128 to denote KeyValue version '2' (what we currently have is KeyValue version
'1').
+ We'd adjust our comparators so they ignored these upper bits in Type so we
can compare Type 1 and Type 2 KeyValues.
+ A total cluster restart will be needed. We'll up the RPC version to be sure.
+ Post-restart, writes will be version 2 KeyValues.
+ Reads may be a mix of version 2 and version 1 KeyValues while regions are
made of a mix of old and new style HFiles. This should be fine as long as
comparators work properly witha mix of version 1 and version 2 KVs.
+ We're thinking the new insertion sequence id will be fixed-length long. A
vint/vlong will make for an awkward parse.
> TestAcidGuarantee broken on trunk
> ----------------------------------
>
> Key: HBASE-2856
> URL: https://issues.apache.org/jira/browse/HBASE-2856
> Project: HBase
> Issue Type: Bug
> Affects Versions: 0.89.20100621
> Reporter: ryan rawson
> Assignee: stack
> Priority: Blocker
> Fix For: 0.92.0
>
>
> TestAcidGuarantee has a test whereby it attempts to read a number of columns
> from a row, and every so often the first column of N is different, when it
> should be the same. This is a bug deep inside the scanner whereby the first
> peek() of a row is done at time T then the rest of the read is done at T+1
> after a flush, thus the memstoreTS data is lost, and previously 'uncommitted'
> data becomes committed and flushed to disk.
> One possible solution is to introduce the memstoreTS (or similarly equivalent
> value) to the HFile thus allowing us to preserve read consistency past
> flushes. Another solution involves fixing the scanners so that peek() is not
> destructive (and thus might return different things at different times alas).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
