[
https://issues.apache.org/jira/browse/HBASE-19568?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16299188#comment-16299188
]
Josh Elser commented on HBASE-19568:
------------------------------------
{code}
+ LOG.debug("Added region procedure manager: " +
regionProcedureClass+"\nAdded region observer: " +
+ regionObserverClass);
{code}
Can you change the newline to a ". " (period and space) just to keep this all
on one line (helps in grepping/parsing logs)
{code}
+ private void updateFileLists(List<String> activeFiles, List<String>
archiveFiles)
+ throws IOException {
+ FileSystem fs = FileSystem.get(conf);
+ List<String> newlyArchived = new ArrayList<String>();
{code}
Should a FileSystem instance be a member to IncrementalTableBackupClient? Or,
pass it in as an argument from {{copyBulkLoadedFiles(..)}}.
{code}
+ private List<Path> getFilesRecursively(String fileBackupDir) throws
IllegalArgumentException,
+ IOException
+ {
+ FileSystem fs = FileSystem.get((new Path(fileBackupDir)).toUri(),
+ new Configuration());
{code}
How about passing in the {{Configuration}} instead of creating a new one? You
have one in {{restoreImages}} already.
{code}
+ RemoteIterator<LocatedFileStatus> it = fs.listFiles(new
Path(fileBackupDir), true);
+ while (it.hasNext()) {
+ Path p = it.next().getPath();
+ if (HFile.isHFileFormat(fs, p)) {
+ list.add(p);
+ }
+ }
{code}
Suggest {{FileSystem.listStatus(Path, PathFilter)}} might be a bit more "terse".
Nice test addition to catch this!
bq. Is the fix mostly done in copyBulkLoadedFiles() ?
Agreed -- any chance I can convince you to move some of the cleanup-related
things (looks mostly like the improvements from the stalled/blocked FT patch
from others)? This seems like a really simple bug to fix, but it gets lost in
the renames. Would be happy to commit the cleanup stuff immediately (given that
was already reviewed).
> Restore of HBase table using incremental backup doesn't restore rows from an
> earlier incremental backup
> -------------------------------------------------------------------------------------------------------
>
> Key: HBASE-19568
> URL: https://issues.apache.org/jira/browse/HBASE-19568
> Project: HBase
> Issue Type: Bug
> Reporter: Romil Choksi
> Assignee: Vladimir Rodionov
> Attachments: HBASE-19568-v1.patch
>
>
> Credits to [~romil.choksi]
> Restore of bulk-loaded HBase table doesn't restore deleted rows
> Steps:
> Create usertable and insert a few rows in it
> Create full backup
> Bulk load into usertable, and create first incremental backup
> Bulk load into usertable again, and create second incremental backup
> Delete row each from initial insert, first bulk load and second bulk load
> Restore usertable using second incremental backup
> Verify if each of the deleted rows has been restored
> On restore using second incremental backup id, the test failed as all of the
> rows from first bulk load were not available. Data from initial insertion
> (full backup) and second bulk load were only available.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
