stack commented on HBASE-19166:

bq.  When the cluster is a mix of HBase 1 and 2 RSs (upgrade in progress) and 
one 2.0 RS crashed and the WAL split is been done by a 1.x server? Am I missing 

You are not missing anything. My thought is Master will put up the WAL for 
splitting, the hbase1 RS will grab it and try to split, fail because it is 
hbase2... and this will go on until a hbase2 RS grabs the WAL. Meantime, we'll 
be adding more RS. I think that will work. We need to spend time on it.

> Add translation for handling hbase.regionserver.wal.WALEdit
> -----------------------------------------------------------
>                 Key: HBASE-19166
>                 URL: https://issues.apache.org/jira/browse/HBASE-19166
>             Project: HBase
>          Issue Type: Bug
>          Components: wal
>            Reporter: Ted Yu
>            Assignee: stack
>            Priority: Blocker
>             Fix For: 2.0.0
> For hlog generated by 1.x, using WALPlayer from hbase2 would result in:
> {code}
> 2017-11-02 21:22:40,907 INFO  [main] mapreduce.Job: Task Id : 
> attempt_1509641483571_0003_m_000000_0, Status : FAILED
> Error: java.lang.ClassCastException: 
> org.apache.hadoop.hbase.regionserver.wal.WALEdit cannot be cast to 
> org.apache.hadoop.hbase.wal.WALEdit
>         at 
> org.apache.hadoop.hbase.mapreduce.WALPlayer$WALCellMapper.map(WALPlayer.java:143)
>         at org.apache.hadoop.mapreduce.Mapper.run(Mapper.java:146)
>         at org.apache.hadoop.mapred.MapTask.runNewMapper(MapTask.java:793)
> {code}
> HBASE-16479 relocated WALEdit.
> Chatting with Enis, he mentioned adding translation for handling 
> hbase.regionserver.wal.WALEdit
> This way, WAL from 1.x can be recognized by hbase-2

This message was sent by Atlassian JIRA

Reply via email to