+1 (non-binding)
Verified RC0 on macOS Intel x86_64, OpenJDK 21.0.9.
Summary of checks (all on the hadoop3 binary tarball / source tarball
from dist.apache.org/dev/hbase/2.6.6RC0/):
== ./dev-support/hbase-vote.sh ==
Ran with -Dhadoop.profile=3.0 -DskipTests:
* Signature : ok
* Checksum : ok
* Rat check : ok
* Built from source : ok (mvn clean install)
* Test phase : skipped (covered separately below)
== Unit tests (runSmallTests) ==
mvn clean package -P runSmallTests -Dhadoop.profile=3.0 \
-Dsurefire.rerunFailingTestsCount=3
Result: 1722 / 1724 passed.
The 2 errors were both in TestMutualTlsClientSideNonLocalhost
(SSLHandshakeException, "Failed to verify both host address and host
name" against the machine's LAN IP). Local issue IG.
== Integration tests on a pseudo-distributed local cluster ==
Setup: 1 master + 3 region servers + 1 ZooKeeper, all separate JVMs
on one machine, file:// rootdir (no HDFS). hbase.master.hostname /
hbase.regionserver.hostname / hbase.zookeeper.quorum pinned to
localhost. JDK21 module-opens flags configured via hbase-env.sh.
* LoadTestTool (LTT) - 1M keys, 10 writers, 1KB values:
write phase: ok
read phase (100% verify): ok
* IntegrationTestBigLinkedList - 1M nodes, 1 mapper, 1 reducer:
Generate + Verify completed.
"Verify finished with success. Total nodes=1000000"
VerifyCounts: REFERENCED=1000000, no UNDEFINED / UNREFERENCED /
EXTRAREFERENCES.
* hbase shell smoke - create / put / get / scan / count /
KeyOnlyFilter / flush / major_compact / disable / drop: all ok.
== Manual chaos test on the same cluster ==
LTT mixed workload (5 writers, ~50% reads, 5M-key target) with a
script killing/restarting one of the secondary region servers every
60s for ~10 minutes during the workload (kept the primary RS alive
to avoid total outage). Rolling chaos exercised meta reassignment,
WAL split, and RS rejoin paths multiple times.
Result:
* Writes: 4,983,391 attempted(as per last logs, actually there were
more), 0 failed
* Actual row count via shell `count`: 4,999,666 (no data loss)
* 11 transient client-side "FAILED FOR cells=NONE" errors clustered
in a single 1-second window immediately after one of the RS
rejoin events; LTT's reader-thread aborter triggered at its
10-error threshold. Region relocation fallout, not an RC defect.
* Master log shows clean RS-dead detection (~30s ZK session
expiry), WAL splits, and region reassignments across all chaos
cycles. No stuck regions; no master crashes.
LGTM for the bits I checked.
Thanks Andrew.
Regard
Umesh Kumar
On Fri, Jun 5, 2026 at 8:12 PM 张铎(Duo Zhang) <[email protected]> wrote:
> +1, binding
>
> Checked sigs and sums: Matched
> Rat check: Passed
> LICENSE and NOTICE: In place
> Compatibility report: We added a method in RegionMetrics interface, should
> be OK
> Built from source:
> mvn clean install -DskipTests -Dhadoop.profile=3.0
> Passed
> Run UTs:
> mvn -Dsurefire.firstPartForkCount=1C
> -Dsurefire.secondPartForkCount=1C -Dsurefire.rerunFailingTestsCount=3
> -Dhadoop.profile=3.0 test -PrunAllTests -fn | tee test.log
> TestMutualTlsClientSideNonLocalhost failed but should be my env problem.
> Start a mini cluster:
> Succeeded.
> Check logs and web pages, nothing unusual
> Run these shell commands, looks fine
> list
> create 't','f'
> put 't','r','f:q','value'
> get 't','r'
> list
> disable 't'
> drop 't'
> list
> Run LTT to load 10000 rows:
> ./bin/hbase ltt -write 4:1024 -num_keys 10000
> Passed
> Run these shell commands, looks fine
> count 'cluster_test'
> scan 'cluster_test', {LIMIT=>10}
> scan 'cluster_test', {FILTER =>
> org.apache.hadoop.hbase.filter.KeyOnlyFilter.new(true)}
> flush 'cluster_test'
> major_compact 'cluster_test'
> disable 'cluster_test'
> drop 'cluster_test'
>
> Peter Somogyi <[email protected]> 于2026年6月5日周五 21:59写道：
> >
> > +1 (binding)
> >
> > hbase-vote.sh:
> > * Signature: ok
> > * Checksum : ok
> > * Rat check (17.0.19): ok
> > - mvn clean apache-rat:check -D hadoop.profile=3.0
> > * Built from source (17.0.19): ok
> > - mvn clean install -D hadoop.profile=3.0 -DskipTests
> > * Unit tests pass (17.0.19): ok
> > - mvn clean package -P runAllTests -D hadoop.profile=3.0
> > -Dsurefire.rerunFailingTestsCount=3
> >
> > Started standalone mode: basic shell operations, the HBase UI and
> > spot-checked the log file.
> >
> > Thanks for the release, Andrew!
> >
> > On Fri, Jun 5, 2026 at 12:20 AM Andrew Purtell <[email protected]>
> wrote:
> >
> > > Please vote on this Apache HBase release candidate, HBase 2.6.6RC0
> > >
> > > The VOTE will remain open for at least 72 hours.
> > >
> > > [ ] +1 Release this package as Apache HBase 2.6.6
> > > [ ] -1 Do not release this package because ...
> > >
> > > The tag to be voted on is 2.6.6RC0:
> > >
> > > https://github.com/apache/hbase/tree/2.6.6RC0
> > >
> > > This tag currently points to git reference
> > >
> > > d5f63f523c065f963992ea4e74db268d17816631
> > >
> > > The release files, including signatures, digests, as well as CHANGES.md
> > > and RELEASENOTES.md included in this RC can be found at:
> > >
> > > https://dist.apache.org/repos/dist/dev/hbase/2.6.6RC0/
> > >
> > > The compatibility report shows one issue in the RegionMetrics
> interface,
> > > but this seems intentional and additive, so I have allowed it. This may
> > > be a factor in your vote.
> > >
> > >
> > >
> > >
> https://dist.apache.org/repos/dist/dev/hbase/2.6.6RC0/api_compare_2.6.5_to_2.6.6RC0.html
> > >
> > > Maven artifacts are available in a staging repository at:
> > >
> > >
> https://repository.apache.org/content/repositories/orgapachehbase-1615/
> > >
> > > Maven artifacts for hadoop3 are available in a staging repository at:
> > >
> > >
> https://repository.apache.org/content/repositories/orgapachehbase-1617/
> > >
> > > Artifacts were signed with the 0xD5365CCD key which can be found in:
> > >
> > > https://downloads.apache.org/hbase/KEYS
> > >
> > > To learn more about Apache HBase, please see
> > >
> > > http://hbase.apache.org/
> > >
> > > Thanks,
> > > Your HBase Release Manager
> > >
>
