> (2) Key put fails for large files (> 20GB) due to a memory leak in Ratis 3.1.0 ...
Duong & Wei-chiu, Thanks for finding this problem! Agree that we should have a Ratis 3.1.1 release. BTW, "Memory leak" usually means that memory was allocated but not released; see https://en.wikipedia.org/wiki/Memory_leak . In this case, we are not having such a problem. Our problem is unnecessarily using too much memory. Tsz-Wo On Tue, Aug 20, 2024 at 6:20 PM Duong Nguyen <du...@cloudera.com.invalid> wrote: > I also filed https://issues.apache.org/jira/browse/RATIS-2141 to track the > memory leak issue. > > Thanks, > Duong > > On Tue, Aug 20, 2024 at 6:17 PM Duong Nguyen <du...@cloudera.com> wrote: > > > Hi all, > > > > I just started a thread to discuss releasing Ratis 3.1.1 with the fixes > of > > the mentioned issues. > > > > Duong > > > > On Tue, Aug 20, 2024 at 5:30 PM Uma Maheswara Rao Gangumalla < > > umaganguma...@gmail.com> wrote: > > > >> Hi Wei-Chiu, > >> > >> Thank you and Duong for the important update on RC1. > >> > >> @Duong would you be notifying this to Ratis community if they can make a > >> quick release with just above 2 fixes? > >> > >> Regards, > >> Uma > >> > >> > >> On Tue, Aug 20, 2024 at 4:51 PM Wei-Chiu Chuang <weic...@apache.org> > >> wrote: > >> > >>> Hi thanks for the effort, > >>> We are testing the latest Ozone master and Ratis 3.1.0 internally, and > >>> found a few critical issues. > >>> > >>> (1) RATIS-2132 <https://issues.apache.org/jira/browse/RATIS-2132> > which > >>> has > >>> about 10% performance regression penalty. > >>> (2) Key put fails for large files (> 20GB) due to a memory leak in > Ratis > >>> 3.1.0: it was a haft-done feature of RATIS-1931. DataNode could crash > due > >>> to out of memory. > >>> > >>> Both of them can only be fixed in Ratis. > >>> I'd suggest to not use Ratis 3.1.0 in Ozone 1.4.1 release. > >>> > >>> If we can, I'd ask the Ratis community to release Ratis 3.1.1 with the > >>> above two fixes. > >>> > >>> cc: @Duong Nguyen <du...@cloudera.com> who helped root cause the two > >>> issues. > >>> > >>> > >>> > >>> > >>> On Tue, Aug 20, 2024 at 3:31 PM Siyao Meng <si...@apache.org> wrote: > >>> > >>> > +1 (binding) > >>> > > >>> > > >>> > - Verified signatures > >>> > - Verified checksums > >>> > - Checked ./bin/ozone version output from binary tarball > >>> > - Checked ./bin/ozone checknative output from binary tarball > >>> > - rocks_tools_native lib check is missing, filed HDDS-11347 > >>> > <https://issues.apache.org/jira/browse/HDDS-11347>, > >>> non-blocking. > >>> > - Checked source tarball content matched repo tag > ozone-1.4.1-RC1 > >>> > - Built from source (without native libs support) > >>> > - Verified compose/ozone Docker dev cluster boots up correctly > with > >>> 3 > >>> > Ozone datanodes. > >>> > - Verified basic volume, bucket, key creation and deletion works > in > >>> > Docker dev cluster. > >>> > - Volume recursive deletion prompt is incorrect, filed > HDDS-11346 > >>> > <https://issues.apache.org/jira/browse/HDDS-11346>, > >>> non-blocking. > >>> > > >>> > > >>> > -Siyao > >>> > > >>> > On Aug 19, 2024 at 6:39:08 AM, Ayush Saxena <ayush...@gmail.com> > >>> wrote: > >>> > > >>> > > +1 (Binding), some minor stuff which we should fix in next release > >>> > > > >>> > > * Built from source > >>> > > * Verified Checksums > >>> > > * Verified Signatures > >>> > > * All source files have apache header > >>> > > * No code diff b/w the git tag & the contents of src tar > >>> > > (dependency-reduced-pom only in src tar, maybe that ain't required > >>> > > there) > >>> > > * Verified the output of ozone version > >>> > > * Ran some basic shell commands > >>> > > * Checked the NOTICE file: The year is *wrong*, it says 2022, it > >>> > > should be 2024 [1], should correct in next release > >>> > > * The NOTICE file inside the packaged Jars is *wrong*, It mentions > >>> > > *Apache Hadoop* & Copyright since 2006, it should be Apache Ozone, > >>> > > should fix in the next release. > >>> > > It currently prints: > >>> > > ``` > >>> > > Apache Hadoop > >>> > > Copyright 2006 and onwards The Apache Software Foundation. > >>> > > . > >>> > > . > >>> > > Hadoop Yarn Server Web Proxy uses the BouncyCastle Java > >>> > > cryptography APIs written by the Legion of the Bouncy Castle Inc. > >>> > > > >>> > > ``` > >>> > > Can try something like to validate: > >>> > > jar xf share/ozone/lib/ozone-client-1.4.1.jar META-INF/NOTICE.txt > >>> > > cat META-INF/NOTICE.txt > >>> > > > >>> > > Thanx Xi Chen for driving the release, Good Luck!!! > >>> > > > >>> > > -Ayush > >>> > > > >>> > > [1] > >>> > > https://github.com/apache/ozone/blob/ozone-1.4.1-RC1/NOTICE.txt#L1-L2 > >>> > > > >>> > > On Mon, 19 Aug 2024 at 11:20, Sammi Chen <sammic...@apache.org> > >>> wrote: > >>> > > > >>> > > > >>> > > +1 (binding) > >>> > > > >>> > > > >>> > > * Verified the signature and checksums > >>> > > > >>> > > * Verified tag > >>> > > > >>> > > * Build from source > >>> > > > >>> > > * Run ozonesecure acceptance test > >>> > > > >>> > > * Start a cluster using bin package > >>> > > > >>> > > * Run freon rk command with data verification > >>> > > > >>> > > * Verified information displayed on Recon UI, for both empty > cluster > >>> and > >>> > > > >>> > > cluster with data > >>> > > > >>> > > Sammi > >>> > > > >>> > > > >>> > > On Fri, 16 Aug 2024 at 13:13, mrchenx <mrch...@126.com> wrote: > >>> > > > >>> > > > >>> > > > Dear Ozone Devs, As discussed in the last email, I am calling > >>> for a > >>> > > > >>> > > > vote on Apache Ozone 1.4.1 RC1. > >>> > > > >>> > > > We have released 1.4.0 on Jan 19th. Now there are 177 new > >>> commits > >>> > > > >>> > > > already landed on 1.4.1 branch, Includes Ratis upgrade (upgrade > to > >>> > Ratis > >>> > > > >>> > > > 3.1.0), some bug fixes, as well as performance optimizations, and > >>> some > >>> > > > >>> > > > necessary dependencies. I am calling for a vote on Apache > Ozone > >>> > 1.4.1 > >>> > > > >>> > > > RC1. - The RC1 tag can be found on Github at: > >>> > > > >>> > > > - > >>> https://github.com/apache/ozone/releases/tag/ozone-1.4.1-RC1 > >>> > > > >>> > > > - 177 Jiras were cherry-pick for ozone-1.4.1 > >>> > > > >>> > > > - > >>> > > > >>> > > > > >>> > > > >>> > > >>> > https://issues.apache.org/jira/issues/?jql=project%20%3D%20HDDS%20AND%20fixVersion%20%3D%201.4.1 > >>> > > > >>> > > > - The source and binary tarballs can be found at: > >>> > > > >>> > > > - > https://dist.apache.org/repos/dist/dev/ozone/1.4.1-rc1/ > >>> > > > >>> > > > - Maven artifacts are staged at: > >>> > > > >>> > > > - > >>> > > > >>> > > > > >>> https://repository.apache.org/content/repositories/orgapacheozone-1024 > >>> > > > >>> > > > - The public key used to sign the artifacts can be found at: > >>> > > > >>> > > > - https://dist.apache.org/repos/dist/release/ozone/KEYS > >>> > > > >>> > > > - The fingerprint of the key used to sign the artifacts is: > >>> > > > >>> > > > - 0D8C19F5514E2786007936F758C87003FF9A1A38 > >>> > > > >>> > > > The vote will run for 7 days, ending on Aug 23th 2024 at 13:10 > >>> pm > >>> > > UTC+8. > >>> > > > >>> > > > > >>> > > > >>> > > > Thanks > >>> > > > >>> > > > > >>> > > > >>> > > > Xi Chen > >>> > > > >>> > > > >>> > > > --------------------------------------------------------------------- > >>> > > To unsubscribe, e-mail: dev-unsubscr...@ozone.apache.org > >>> > > For additional commands, e-mail: dev-h...@ozone.apache.org > >>> > > > >>> > > > >>> > > >>> > >> >
