[
https://issues.apache.org/jira/browse/HBASE-28078?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17764284#comment-17764284
]
Bryan Beaudreault commented on HBASE-28078:
-------------------------------------------
Yep, I agree this looks inaccurate. As you said, the actual incremental backup
process converts WAL to HFile. On restore, it just pulls down all HFiles of the
last full backup and any incremental backups.
> Update tech details of backup
> -----------------------------
>
> Key: HBASE-28078
> URL: https://issues.apache.org/jira/browse/HBASE-28078
> Project: HBase
> Issue Type: Task
> Reporter: Dieter De Paepe
> Priority: Minor
> Original Estimate: 2h
> Remaining Estimate: 2h
>
> The [technical
> details|https://hbase.apache.org/book.html#br.technical.details] of HBase
> backup mention that the backup works by copying WALs which are converted to
> HFiles during the restore process.
> However, the code seems to imply that WALs are already converted to HFiles
> during the backup process. (See
> IncrementalTableBackupClient#convertWALsToHFiles.) I suspect the docs still
> describe the old behavior, which (I suspect) was changed in HBase-14141.
> Knowing this also confirms that bulk-imported HFiles are also backed-up using
> the same mechanism (see IncrementalTableBackupClient#handleBulkLoad) -
> something I managed to test with experimentation, but wasn't explained in the
> backup docs.
> In short, this section in the docs requires an update.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
