[
https://issues.apache.org/jira/browse/KUDU-3562?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17832934#comment-17832934
]
ASF subversion and git services commented on KUDU-3562:
-------------------------------------------------------
Commit e54c80cf9ba129b943eed6af3266899753b1b4ac in kudu's branch
refs/heads/master from Alexey Serbin
[ https://gitbox.apache.org/repos/asf?p=kudu.git;h=e54c80cf9 ]
KUDU-3562 fix integer overflow in available space metrics
This patch addresses KUDU-3562. The problem was in implicit casting
of 64-bit integers down to 32-bit ones when introducing a helper
variable in CalculateAvailableSpace(). A new test scenario is added
to cover the fixed issue and catch regressions in future, if any.
In addition, this patch corrects the type of the related metrics in
disk_failure-itest.cc. It also clarifies on the description and tags
of the --fs_{data_dirs,wal_dir}_available_space_cache_seconds flags.
Change-Id: I974aea822626e4648886388c0de3741ac459f2ec
Reviewed-on: http://gerrit.cloudera.org:8080/21227
Reviewed-by: Abhishek Chennaka <[email protected]>
Tested-by: Abhishek Chennaka <[email protected]>
Reviewed-by: Mahesh Reddy <[email protected]>
> Inconsistency in 'available space' metrics when reserving more than 2GiB
> ------------------------------------------------------------------------
>
> Key: KUDU-3562
> URL: https://issues.apache.org/jira/browse/KUDU-3562
> Project: Kudu
> Issue Type: Bug
> Components: master, tserver
> Affects Versions: 1.16.0, 1.17.0
> Reporter: Alexey Serbin
> Assignee: Alexey Serbin
> Priority: Major
>
> When setting {{\-\-fs_data_dirs_reserved_bytes}} and
> {{\-\-fs_wal_dir_reserved_bytes}} flags to values greater than 2GiBytes, the
> {{wal_dir_space_available_bytes}} and {{data_dirs_space_available_bytes}}
> reports incorrect numbers for both Kudu masters and tablet servers. The
> metrics report much more space than actually available.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)