[
https://issues.apache.org/jira/browse/HBASE-5074?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13195952#comment-13195952
]
dhruba borthakur commented on HBASE-5074:
-----------------------------------------
Thanks Ted/Zhihong for the review comments.
ted: The thinking is that HFileSystem.readfs should be used only by StoreFiles
for reading hfile. That is the reason that this code path uses readfs. This is
the only place we want to avoid using hdfs checksums. All other code paths are
unchanged.
Zhihong: If hbase checksum validation fails once, I switch back to using
hdfs-level checksums for that instance of the Reader. For each block that have
hbase-checksums mismatch we retry the operation, thus it actually *doubles* the
iops. I was trying to avoid the scenario where most hbase level checksums fail,
and each io is retried twice. But if people feel otherwise, I can set the
useHBaseChecksum after a few successful ios.
I will log the fact the hdfs checksum verification worked and will also add
metrics counters to record these events (next version of the patch).
> support checksums in HBase block cache
> --------------------------------------
>
> Key: HBASE-5074
> URL: https://issues.apache.org/jira/browse/HBASE-5074
> Project: HBase
> Issue Type: Improvement
> Components: regionserver
> Reporter: dhruba borthakur
> Assignee: dhruba borthakur
> Attachments: D1521.1.patch, D1521.1.patch
>
>
> The current implementation of HDFS stores the data in one block file and the
> metadata(checksum) in another block file. This means that every read into the
> HBase block cache actually consumes two disk iops, one to the datafile and
> one to the checksum file. This is a major problem for scaling HBase, because
> HBase is usually bottlenecked on the number of random disk iops that the
> storage-hardware offers.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
