[
https://issues.apache.org/jira/browse/HDFS-6038?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jing Zhao updated HDFS-6038:
----------------------------
Description:
In HA setup, the JNs receive edit logs (blob) from the NN and write into edit
log files. In order to write well-formed edit log files, the JNs prepend a
header for each edit log file. The problem is that the JN hard-codes the
version (i.e., {{NameNodeLayoutVersion}} in the edit log, therefore it
generates incorrect edit logs when the newer release bumps the
{{NameNodeLayoutVersion}} during rolling upgrade.
In the meanwhile, currently JN tries to decode the in-progress editlog segment
in order to know the last txid in the segment. In the rolling upgrade scenario,
the JN with the old software may not be able to correctly decode the editlog
generated by the new software.
This jira makes the following changes to allow JN to handle editlog produced by
software with future layoutversion:
1. Change the NN--JN startLogSegment RPC signature and let NN specify the
layoutversion for the new editlog segment.
2. Persist a length field for each editlog op to indicate the total length of
the op. Instead of calling EditLogFileInputStream#validateEditLog to get the
last txid of an in-progress editlog segment, a new method scanEditLog is added
and used by JN which does not decode each editlog op but uses the length to
quickly jump to the next op.
was:
In HA setup, the JNs receive edit logs (blob) from the NN and write into edit
log files. In order to write well-formed edit log files, the JNs prepend a
header for each edit log file.
The problem is that the JN hard-codes the version (i.e.,
{{NameNodeLayoutVersion}} in the edit log, therefore it generates incorrect
edit logs when the newer release bumps the {{NameNodeLayoutVersion}} during
rolling upgrade.
> JournalNode hardcodes NameNodeLayoutVersion in the edit log file
> ----------------------------------------------------------------
>
> Key: HDFS-6038
> URL: https://issues.apache.org/jira/browse/HDFS-6038
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: journal-node, namenode
> Reporter: Haohui Mai
> Assignee: Jing Zhao
> Attachments: HDFS-6038.000.patch, HDFS-6038.001.patch,
> HDFS-6038.002.patch, HDFS-6038.003.patch, HDFS-6038.004.patch,
> HDFS-6038.005.patch, HDFS-6038.006.patch, HDFS-6038.007.patch, editsStored
>
>
> In HA setup, the JNs receive edit logs (blob) from the NN and write into edit
> log files. In order to write well-formed edit log files, the JNs prepend a
> header for each edit log file. The problem is that the JN hard-codes the
> version (i.e., {{NameNodeLayoutVersion}} in the edit log, therefore it
> generates incorrect edit logs when the newer release bumps the
> {{NameNodeLayoutVersion}} during rolling upgrade.
> In the meanwhile, currently JN tries to decode the in-progress editlog
> segment in order to know the last txid in the segment. In the rolling upgrade
> scenario, the JN with the old software may not be able to correctly decode
> the editlog generated by the new software.
> This jira makes the following changes to allow JN to handle editlog produced
> by software with future layoutversion:
> 1. Change the NN--JN startLogSegment RPC signature and let NN specify the
> layoutversion for the new editlog segment.
> 2. Persist a length field for each editlog op to indicate the total length of
> the op. Instead of calling EditLogFileInputStream#validateEditLog to get the
> last txid of an in-progress editlog segment, a new method scanEditLog is
> added and used by JN which does not decode each editlog op but uses the
> length to quickly jump to the next op.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
