[
https://issues.apache.org/jira/browse/HBASE-20447?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16441782#comment-16441782
]
Ted Yu commented on HBASE-20447:
--------------------------------
{code}
+ * if not, throw an exception. If they are the same without the
nextBlockMetadata, return the comparison.
{code}
RuntimeException is thrown. Have you considered throwing IOE ?
{code}
+ LOG.warn("Cached block contents differ, trying to just compare the block
contents " +
+ "without the next block. CacheKey: " + cacheKey);
{code}
Is there anything admin can do after seeing the above log ?
{code}
+ LOG.warn("Cached block contents differ by nextBlockOnDiskSize.
Keeping cached block.");
+ return;
+ } else {
+ LOG.warn("Cached block contents differ by nextBlockOnDiskSize.
Caching new block.");
{code}
The first part of the log is the same for both cases. Is it possible to make
the log clearer as to why the decision of caching is made ?
{code}
+ if (includeNextBlockMetadata) {
+ destination.putInt(this.nextBlockOnDiskSize);
{code}
The flag only guards one field. Would includeNextBlockOnDiskSize be better name
for the parameter ?
> Only fail cacheBlock if block collisions aren't related to next block metadata
> ------------------------------------------------------------------------------
>
> Key: HBASE-20447
> URL: https://issues.apache.org/jira/browse/HBASE-20447
> Project: HBase
> Issue Type: Bug
> Components: BlockCache, BucketCache
> Affects Versions: 1.4.3, 2.0.0
> Reporter: Zach York
> Assignee: Zach York
> Priority: Major
> Attachments: HBASE-20447.branch-1.001.patch
>
>
> This is the issue I was originally having here:
> [http://mail-archives.apache.org/mod_mbox/hbase-dev/201802.mbox/%3CCAN+qs_Pav=md_aoj4xji+kcnetubg2xou2ntxv1g6m8-5vn...@mail.gmail.com%3E]
>
> When we pread, we don't force the read to read all of the next block header.
> However, when we get into a race condition where two opener threads try to
> cache the same block and one thread read all of the next block header and the
> other one didn't, it will fail the open process. This is especially important
> in a splitting case where it will potentially fail the split process.
> Instead, in the caches, we should only fail if the required blocks are
> different.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
