[
https://issues.apache.org/jira/browse/HDFS-13863?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16594780#comment-16594780
]
Yiqun Lin commented on HDFS-13863:
----------------------------------
Yes, [~ferhui], the exception message can tell us why we fall back. The main
point from me is that we should let users know the scenery, here is for placing
the block on a transient volume. I mean we should mention this. Though this
again, use "Insufficient space error for ..." looks better.
> FsDatasetImpl should log DiskOutOfSpaceException
> ------------------------------------------------
>
> Key: HDFS-13863
> URL: https://issues.apache.org/jira/browse/HDFS-13863
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: hdfs
> Affects Versions: 3.1.0, 2.9.1, 3.0.3
> Reporter: Fei Hui
> Assignee: Fei Hui
> Priority: Major
> Attachments: HDFS-13863.001.patch, HDFS-13863.002.patch
>
>
> The code in function *createRbw* as follow
> {code:java}
> try {
> // First try to place the block on a transient volume.
> ref = volumes.getNextTransientVolume(b.getNumBytes());
> datanode.getMetrics().incrRamDiskBlocksWrite();
> } catch (DiskOutOfSpaceException de) {
> // Ignore the exception since we just fall back to persistent
> storage.
> } finally {
> if (ref == null) {
> cacheManager.release(b.getNumBytes());
> }
> }
> {code}
> I think we should log the exception because it took me long time to resolve
> problems, and maybe others face the same problems.
> When i test ram_disk, i found no data was written into randomdisk. I debug,
> deep into the source code, and found that randomdisk size was less than
> reserved space. I think if message was logged, i would resolve the problem
> quickly.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
