[
https://issues.apache.org/jira/browse/CASSANALYTICS-98?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Yifan Cai updated CASSANALYTICS-98:
-----------------------------------
Reviewers: Doug Rohrer, Yifan Cai
Status: Review In Progress (was: Patch Available)
> Fix SSTable descriptor mismatch preventing newly produced SSTables from
> being uploaded
> ----------------------------------------------------------------------------------------
>
> Key: CASSANALYTICS-98
> URL: https://issues.apache.org/jira/browse/CASSANALYTICS-98
> Project: Apache Cassandra Analytics
> Issue Type: Bug
> Components: Writer
> Reporter: Yifan Cai
> Assignee: Yifan Cai
> Priority: Normal
> Time Spent: 40m
> Remaining Estimate: 0h
>
> The bug was in SSTables.getSSTableBaseName() which included a trailing dash
> (e.g., "oa-1-big-") while SSTable descriptors created during production
> didn't have the trailing dash (e.g., "oa-1-big"). This mismatch caused the
> filter in SortedSSTableWriter.prepareSStablesToSend() to exclude all newly
> produced SSTables from upload, as the descriptors never matched.
> The fix removes the trailing dash by changing lastIndexOf("-") + 1 to
> lastIndexOf("-") and adds corresponding logic to avoid double-hashing
> already-streamed SSTables.
> The bug does not cause data loss. All produced sstables are still uploaded
> when the writer is closed. The bugs is causing the delay of sending those
> files, as they could have been sent when they are just produced. I believe
> the bug was introduced in CASSANALYTICS-5.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
