[
https://issues.apache.org/jira/browse/HADOOP-1134?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12483275
]
Doug Cutting commented on HADOOP-1134:
--------------------------------------
> I thought since DN has complete control of how checksums are done, it could
> as well decide size various parameters for it.
The datanode primarily only has control of how checksums are stored. Checksums
are created and consumed primarily by the client. So my instinct would be to
have the client determine checksum creation parameters. This is also simpler,
as you observed.
> Initially DN detects corrupted blocks when they are read for any reason.
To be clear: I think the first priority should be to detect corruption in the
client. Then, only subsequently might we add checking on the datanode itself.
Checks on the datanode might only happen when a client-side check fails, to
determine whether the problem is on the disk or happened in transit, and
perhaps during periodic scans. The most important check however is the
client's check.
> Block level CRCs in HDFS
> ------------------------
>
> Key: HADOOP-1134
> URL: https://issues.apache.org/jira/browse/HADOOP-1134
> Project: Hadoop
> Issue Type: New Feature
> Components: dfs
> Reporter: Raghu Angadi
> Assigned To: Raghu Angadi
>
> Currently CRCs are handled at FileSystem level and are transparent to core
> HDFS. See recent improvement HADOOP-928 ( that can add checksums to a given
> filesystem ) regd more about it. Though this served us well there a few
> disadvantages :
> 1) This doubles namespace in HDFS ( or other filesystem implementations ). In
> many cases, it nearly doubles the number of blocks. Taking namenode out of
> CRCs would nearly double namespace performance both in terms of CPU and
> memory.
> 2) Since CRCs are transparent to HDFS, it can not actively detect corrupted
> blocks. With block level CRCs, Datanode can periodically verify the checksums
> and report corruptions to namnode such that name replicas can be created.
> We propose to have CRCs maintained for all HDFS data in much the same way as
> in GFS. I will update the jira with detailed requirements and design. This
> will include same guarantees provided by current implementation and will
> include a upgrade of current data.
>
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
