[
https://issues.apache.org/jira/browse/HTRACE-304?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15023088#comment-15023088
]
Colin Patrick McCabe commented on HTRACE-304:
---------------------------------------------
Thanks for trying out the test, [~iwasakims]. It seems that the bug was masked
by the fact that spans were divided between two different leveldb databases,
and the "one skip" logic was done on a per-leveldb basis. I have changed the
test to use only one leveldb instance (i.e. shard) and now it fails properly
prior to the fix, and succeeds after it. Check out 002.
> htraced: fix bug with GREATER_THAN queries
> ------------------------------------------
>
> Key: HTRACE-304
> URL: https://issues.apache.org/jira/browse/HTRACE-304
> Project: HTrace
> Issue Type: Bug
> Components: htraced
> Affects Versions: 4.0
> Reporter: Colin Patrick McCabe
> Assignee: Colin Patrick McCabe
> Attachments: HTRACE-304.001.patch, HTRACE-304.002.patch
>
>
> htraced has a bug with greater_than queries.
> Leveldb exposes a sorted map interface. It supports seeking to a specific
> key, but not to a position "greater than" a specific key. The code right now
> for greater_than assumes that if after the leveldb seek, it sees an
> equivalent key, it only needs to do one step forward to get to the keys which
> it should be able to return. However, it there are multiple "equal" entries,
> greater_than may have to do multiple seeks.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
