[
https://issues.apache.org/jira/browse/HDFS-12979?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16835157#comment-16835157
]
Plamen Jeliazkov commented on HDFS-12979:
-----------------------------------------
Hey Chen, Erik,
Good work on this so far. I have some questions / suggestions.
If we are going to have multiple SbNNs sending checkpoints, can we allow for
some new configuration parameter logic to deal with which SbNN sends to whom?
My concern is that as it is in v.003 we would have to maintain separate
configuration files for different SbNNs in order to achieve this due to:
{code:java}
// Look up the active node's address
List<Configuration> confForActive = HAUtil.getConfForOtherNodes(conf);
activeNNAddresses = new ArrayList<URL>(confForActive.size());
{code}
It would be nice for both of them to have the same configuration and just know
which targets they send to (and that way can validate they don't have
intersecting targets too).
The FIndBugs comment about using a map / set of URLs is important here because
hashCode and equals of URL does domain name resolution. Making it expensive.
Look at URLStreamHandler's hashCode(URL). It does not do parameter Object equal
bypassing either. Probably easier to just use String representation of the URL,
or a List of Pairs.
Some nits:
`private void doCheckpoint(boolean stayQuiet)` -- does stayQuiet make sense
here? Seems more it's a boolean for skipping checkpoint transfer. Maybe
skipUploads?
I think Erik pointed this out earlier, but we should probably rename
`activeNNAddresses` now inside StandbyCheckpointer. Maybe
checkpointUploadTargets?
Thanks for all your work.
> StandbyNode should upload FsImage to ObserverNode after checkpointing.
> ----------------------------------------------------------------------
>
> Key: HDFS-12979
> URL: https://issues.apache.org/jira/browse/HDFS-12979
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: hdfs
> Reporter: Konstantin Shvachko
> Assignee: Chen Liang
> Priority: Major
> Attachments: HDFS-12979.001.patch, HDFS-12979.002.patch,
> HDFS-12979.003.patch
>
>
> ObserverNode does not create checkpoints. So it's fsimage file can get very
> old making bootstrap of ObserverNode too long. A StandbyNode should copy
> latest fsimage to ObserverNode(s) along with ANN.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
