Apache
/
hbase
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

HBASE-8770 deletes and puts with the same ts should be resolved according to mvcc/seqNum

This commit is contained in:
Michael Stack 2018-03-01 23:51:08 -08:00
parent 1f5e93a8f8
commit 1bafe4a028
1 changed files with 35 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

35
src/main/asciidoc/_chapters/datamodel.adoc
View File

@ -343,6 +343,7 @@ In particular:
Below we describe how the version dimension in HBase currently works.
See link:https://issues.apache.org/jira/browse/HBASE-2406[HBASE-2406] for discussion of HBase versions. link:https://www.ngdata.com/bending-time-in-hbase/[Bending time in HBase] makes for a good read on the version, or time, dimension in HBase.
It has more detail on versioning than is provided here.
As of this writing, the limitation _Overwriting values at existing timestamps_ mentioned in the article no longer holds in HBase.
This section is basically a synopsis of this article by Bruno Dumon.
@ -503,8 +504,42 @@ Otherwise, a delete marker with a timestamp in the future is kept until the majo
NOTE: This behavior represents a fix for an unexpected change that was introduced in HBase 0.94, and was fixed in link:https://issues.apache.org/jira/browse/HBASE-10118[HBASE-10118].
The change has been backported to HBase 0.94 and newer branches.
[[new.version.behavior]]
=== Optional New Version and Delete behavior in HBase-2.0.0
In `hbase-2.0.0`, the operator can specify an alternate version and
delete treatment by setting the column descriptor property
`NEW_VERSION_BEHAVIOR` to true (To set a property on a column family
descriptor, you must first disable the table and then alter the
column family descriptor; see <<cf.keep.deleted>> for an example
of editing an attribute on a column family descriptor).
The 'new version behavior', undoes the limitations listed below
whereby a `Delete` ALWAYS overshadows a `Put` if at the same
location -- i.e. same row, column family, qualifier and timestamp
-- regardless of which arrived first. Version accounting is also
changed as deleted versions are considered toward total version count.
This is done to ensure results are not changed should a major
compaction intercede. See `HBASE-15968` and linked issues for
discussion.
Running with this new configuration currently costs; we factor
the Cell MVCC on every compare so we burn more CPU. The slow
down will depend. In testing we've seen between 0% and 25%
degradation.
If replicating, it is advised that you run with the new
serial replication feature (See `HBASE-9465`; the serial
replication feature did NOT make it into `hbase-2.0.0` but
should arrive in a subsequent hbase-2.x release) as now
the order in which Mutations arrive is a factor.
=== Current Limitations
The below limitations are addressed in hbase-2.0.0. See
the section above, <<new.version.behavior>>.
==== Deletes mask Puts
Deletes mask puts, even puts that happened after the delete was entered.