hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Todd Lipcon (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-4403) DFSClient can infer checksum type when not provided by reading first byte
Date Tue, 15 Jan 2013 22:04:13 GMT

    [ https://issues.apache.org/jira/browse/HDFS-4403?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13554423#comment-13554423

Todd Lipcon commented on HDFS-4403:

bq. The old client code just called getCrcType() without checking hasCrcType(). The old clients
talking to new server with this change, get null pointer exception, if the new server does
not set the crcType

That's not how defaults work in protobuf -- the protobuf maintains a separate flag for "hasFieldX",
which is set to false upon construction even if the field has a default. So, on the server
side, if an optional field isn't explicitly set, it won't serialize the default of that field
to the wire when it's serialized. That is to say, given that the field has always been optional,
the case of a server not setting it will be handled the same regardless of whether or not
it has a default.

To illustrate, I made the following test proto:
message MyTestProto {
  optional string testField = 1 [ default = "hello world" ];

And ran the following code:

    MyTestProto pb = MyTestProto.newBuilder().build();
    System.err.println("has field? " + pb.hasTestField());
    System.err.println("field value: " + pb.getTestField());
    System.err.println("serialized: " + StringUtils.byteToHexString(pb.toByteArray()));


has field? false
field value: hello world

Note how the default value isn't put on the "wire" when the protobuf is serialized to a byte

So, in your example, with the old clients talking to a new server which doesn't set crc type,
the old clients would continue to use whatever default they'd defined locally.
> DFSClient can infer checksum type when not provided by reading first byte
> -------------------------------------------------------------------------
>                 Key: HDFS-4403
>                 URL: https://issues.apache.org/jira/browse/HDFS-4403
>             Project: Hadoop HDFS
>          Issue Type: Bug
>          Components: hdfs-client
>    Affects Versions: 2.0.2-alpha
>            Reporter: Todd Lipcon
>            Assignee: Todd Lipcon
>            Priority: Minor
>         Attachments: hdfs-4403.txt
> HDFS-3177 added the checksum type to OpBlockChecksumResponseProto, but the new protobuf
field is optional, with a default of CRC32. This means that this API, when used against an
older cluster (like earlier 0.23 releases) will falsely return CRC32 even if that cluster
has written files with CRC32C. This can cause issues for distcp, for example.
> Instead of defaulting the protobuf field to CRC32, we can leave it with no default, and
if the OpBlockChecksumResponseProto has no checksum type set, the client can send OP_READ_BLOCK
to read the first byte of the block, then grab the checksum type out of that response (which
has always been present)

This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

View raw message