hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sameer Paranjpye (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HADOOP-1134) Block level CRCs in HDFS
Date Thu, 29 Mar 2007 20:03:25 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-1134?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12485328

Sameer Paranjpye commented on HADOOP-1134:

> Okay, I see your point. If we import only a single replica of the checksums on upgrade
then we'd increase the false-positive rate of checksum errors, right? But for every 
> false-positive we'd still have a 100 real data corruptions, so I'm not sure this is a
big deal.

Yes, but in each of those 100 real data corruptions data can be salvaged by switching to a
valid instance of the block, deleting the corrupt instance and replicating the valid instances.
If we blindly copy checksums we'll introduce errors that cannot be recovered from, period.

> Yes, in some cases. If the corruption happened to a checksum on a datanode, then it does.
If it happened before the data reached the datanode, then it doesn't.

Not in just some cases, but in the overwhelming majority of cases. Data corruption before
data reaches the Datanode would occur either in RAM or during network transmission, the likelihood
of this happening is orders of magnitude lower than 1 out of 3 replicas on disk becoming corrupt.
Disks are flaky, far flakier than RAM or network, and bit rot happens all the time when you
have a large number of them. The percentage of corruptions that would occur before data reaches
the Datanodes is far lower than 1, based on the error rates I'm aware of.

> 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.

View raw message