hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Todd Lipcon (Commented) (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-1975) HA: Support for sharing the namenode state from active to standby.
Date Thu, 17 Nov 2011 22:27:52 GMT

    [ https://issues.apache.org/jira/browse/HDFS-1975?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13152404#comment-13152404

Todd Lipcon commented on HDFS-1975:

- Why is {{getMaxGsInBlockList}} static? seems it could just be a member function of {{BlockListAsLongs}}
- Storage has a javadoc {{@param shouldLock}} but the parameter doesn't seem to be in the
method signature


In {{EditLogFileInputStream}}, why do we need to pass {{isInProgress}} as a boolean? If it's
in-progress, that means we don't know the {{lastTxId}}, so it would be {{INVALID_TXID}}, right?
So we can implement {{isInProgress()}} by just comparing {{lastTxId}}.


- There's a typo "UNITIALIZED" instead of "UNINITIALIZED" in the javadoc in {{FSEditLog}}
- The comment before {{sharedEditsDirs}} in {{FSEditLog}} should be javadoc-style
- In the FSEditLog state machine, how does the transition from OPEN_FOR_READING work when
going active? The javadoc could use a little bit more there (do we go first to CLOSED?)
- Similar to above - would be good to add Preconditions.checkState checks in {{initJournalsForWrite}}
and {{initSharedJournalsForRead}} - it's not obvious what state they should be to avoid orphaning
open streams, etc.
- Assertion in {{logEdit}}: can you add a text error message like {{"bad state: " + state}}
so that if the assertion fails we're left with more useful info?


- Do you really need to make all of the FSN methods VisibleForTesting? We have a class called
NameNodeAdapter which you can probably use to "reach in" without changing visibility. Or,
why not just make a non-HA DFSClient talking to the first NN in the test case?
- {{editLogTailer}} should be defined up higher in the file, no?
- it seems like the {{recoverUnclosedStreams}} should happen just changing to writer mode,
rather than at {{stopReadingEditLogs}} (doesn't seem obvious that this method would mutate
the dir state). Otherwise when we clean-shutdown the standby, it might try to move around
some logs, no?

- Why does {{matchEditLogs}} accept null now? It's called with the result of {{FileUtil.listFiles}}
which never returns null
- A couple spurious changes in FileJournalManager, JournalManager

- the builder method should be {{haEnabled}} not {{setHaEnabled}} to match the pattern of
the other methods

- Need a license on TestEditLogTailer
> HA: Support for sharing the namenode state from active to standby.
> ------------------------------------------------------------------
>                 Key: HDFS-1975
>                 URL: https://issues.apache.org/jira/browse/HDFS-1975
>             Project: Hadoop HDFS
>          Issue Type: Sub-task
>          Components: name-node
>            Reporter: Suresh Srinivas
>            Assignee: Jitendra Nath Pandey
>         Attachments: HDFS-1975-HA.2.patch, HDFS-1975-HA.patch, HDFS-1975-HDFS-1623.patch,
hdfs-1975.txt, hdfs-1975.txt
> To enable hot standby namenode, the standby node must have current information for -
namenode state (image + edits) and block location information. This jira addresses keeping
the namenode state current in the standby node. To do this, the proposed solution in this
jira is to use a shared storage to store the namenode state. 
> Note one could also build an alternative solution by augmenting the backup node. A seperate
jira could explore this.

This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira


View raw message