[
https://issues.apache.org/jira/browse/HBASE-21401?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16702761#comment-16702761
]
Zheng Hu commented on HBASE-21401:
----------------------------------
bq. What I meant is that the our Stream to HDFS is a checksummed stream.
Understood. But I don't think the checksumed stream is enough. if so, then the
HBASE-21379 won't happen, i think. I guess the mis-encoding may happen in
client side or server side. the client or server did not realize that the
encode was wrong, they generate the mis-encoding bytes's checksum and check
again, so did not detect the bug.
bq. Just don't understand why we have to parse the bytes twice – once to check,
and then once to construct KV; can we not integrate the check with the KV
construction so we don't do length parses for row, offsets to type and
timestamp, etc., twice?
I think it's balance between space and time. If parse once, then all the
offsets need to be saved in KeyValue object, may have lots of small object ?
we can see we only have the mininal members in KeyValue class: byte[], offset,
length, seqId..... the parse is cheap as we can see in HBASE-21459, so I think
it's OK.
Thanks.
> Sanity check in BaseDecoder#parseCell
> -------------------------------------
>
> Key: HBASE-21401
> URL: https://issues.apache.org/jira/browse/HBASE-21401
> Project: HBase
> Issue Type: Sub-task
> Components: regionserver
> Reporter: Zheng Hu
> Assignee: Zheng Hu
> Priority: Critical
> Fix For: 3.0.0, 2.2.0, 2.0.3, 2.1.2
>
> Attachments: HBASE-21401.v1.patch, HBASE-21401.v2.patch,
> HBASE-21401.v3.patch, HBASE-21401.v4.patch, HBASE-21401.v4.patch,
> HBASE-21401.v5.patch
>
>
> In KeyValueDecoder & ByteBuffKeyValueDecoder, we pass a byte buffer to
> initialize the Cell without a sanity check (check each field's offset&len
> exceed the byte buffer or not), so ArrayIndexOutOfBoundsException may happen
> when read the cell's fields, such as HBASE-21379, it's hard to debug this
> kind of bug.
> An earlier check will help to find such kind of bugs.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
