HIVE-29477: Introduce TezSession interface and a new implementation for unmanaged, external Tez sessions by abstractdog · Pull Request #6343 · apache/hive

abstractdog · 2026-03-02T14:39:18Z

What changes were proposed in this pull request?

HiveConf: add options for external sessions discovered from Zookeeper
TezSession: new interface for subclasses like TezSessionState, TezExternalSessionState
TezExternalSessionState: the implementation of the new, unmanaged Tez session, where openInternal simply acquired an ApplicationId instead of submitting a Yarn application
ExternalSessionsRegistry: interface for a Tez session registry that can be implemented for specific frameworks and environments, currently ZookeeperExternalSessionsRegistryClient - this interface complies with a simple contract to get + return a Tez session for Hive
AbstractTriggerValidator: top-level class instead of TezSessionPoolSession.AbstractTriggerValidator
TezSessionState -> TezSession: it's changed at many places in the code, because most of the Hive code has nothing to do with the underlying implementation, so with the fact whether those are managed or unmanaged sessions

Why are the changes needed?

Make able HiveServer2 to submit DAGs to external sessions (not managed by HiveServer2 and Yarn).

Does this PR introduce any user-facing change?

Yes.

How was this patch tested?

Unit tests added, e2e tests performed with Tez AM docker image from currently in-progress TEZ-4682.

Aggarwal-Raghav · 2026-03-03T18:06:58Z

ql/src/java/org/apache/hadoop/hive/ql/session/SessionState.java

   */
  private final Map<String, FunctionInfo> currentFunctionsInUse = new HashMap<>();

+  private static ExternalSessionsRegistry externalSessions = null;


can be made as a local variable in start() method

doesn't even need a reference on this registry here, removing it altogether

Aggarwal-Raghav · 2026-03-03T18:29:01Z

ql/src/java/org/apache/hadoop/hive/ql/exec/tez/ZookeeperExternalSessionsRegistryClient.java

+  private final Object lock = new Object();
+  private final int maxAttempts;
+
+  private PathChildrenCache cache;


PathChildrenCache is deprecated. Can you please check if we can use CuratorCache API's

abstractdog · 2026-03-06T12:19:12Z

@Aggarwal-Raghav : thanks for your comments so far: can you please check if this patch can work together with your current testing scenario regarding TEZ-4682, I'm specifically interested whether TEZ-4686 goes away if hive is configured accordingly:

hive.server2.use.external.sessions=true
hive.server2.use.external.sessions.namespace=...
hive.server2.tez.external.sessions.registry.class=org.apache.hadoop.hive.ql.exec.tez.ZookeeperExternalSessionsRegistryClient

Aggarwal-Raghav · 2026-03-07T21:30:19Z

Hi @abstractdog , i tested this patch on my local setup (hadoop, hive, TEZ-4682 branch and Zookeeper - installed via brew) and its working🚀. Ran a basic insert command

Attaching the HS2 logs and tez_am docker logs

tez-am-working.log
working_hs2.log

Few things to note:

Add the following in /etc/hosts/ file. this is necesary for docker to communicate with hdfs running on localhost
127.0.0.1 host.docker.internal
tez-am docker image needs hive-exec jar i.e resorce localization. For now I added it using the plugin directory way i.e. -v "/Users/raghav/Desktop/plugin:/opt/tez/plugins" \ and plugin directory contains only hive-exec jar otherwise .

Vertex vertex_1769280834537_0000_1_01 [Reducer 2] killed/failed due to: INIT_FAILURE] ....
vertex=vertex_1769280834537_0000_1_00 [Map 1], org.apache.tez.dag.api .TezUncheckedException: java. lang.RuntimeException: java.lang.RuntimeException: java.lang.ClassNotFoundException: Class org.apache.ha
doop.hive.ql.io. CombineHiveInputFormat not found

core-site.xml

    <property>
        <name>fs.defaultFS</name>
        <value>hdfs://host.docker.internal:9000</value>
    </property>

hdfs-site.xml Connecting docker AM to datanode was really problematic, took lot of time 😅

    <property>
      <name>dfs.datanode.use.datanode.hostname</name>
      <value>true</value>
    </property>

    <property>
      <name>dfs.datanode.address</name>
      <value>host.docker.internal:9866</value>
    </property>

    <property>
      <name>dfs.datanode.hostname</name>
      <value>host.docker.internal</value>
    </property>

In docker tez-am tez-site.xml (Will update the PR for this as well in TEZ-4682)

    <property>
        <name>dfs.client.use.datanode.hostname</name>
        <value>true</value>
    </property>

Some --add-opens in tez-entrypoint.sh needs to be updated. Basically hive project --add-opens are also required. Will update the PR for TEZ-4682.
hive-site.xml

  <property>
      <name>hive.server2.use.external.sessions</name>
      <value>true</value>
  </property>

  <property>
    <name>hive.server2.tez.external.sessions.namespace</name>
    <value>/tez-external-sessions</value>
  </property>

  <property>
      <name>hive.server2.tez.external.sessions.registry.class</name>
      <value>org.apache.hadoop.hive.ql.exec.tez.ZookeeperExternalSessionsRegistryClient</value>
  </property>

  <property>
    <name>hive.zookeeper.quorum</name>
    <value>localhost:2181</value>
  </property>

NOTE/IMPORTANT: There was a flaky behaviour observed with tez-conf.pb in tez-staging directory in hdfs. It was throwing error for No file found. Not sure why it was occuring but it was fixed after some time automatically and not facing it again. Will post here if I face it again.

abstractdog · 2026-03-09T12:49:17Z

hive.server2.tez.external.sessions.namespace

Hi @abstractdog , i tested this patch on my local setup (hadoop, hive, TEZ-4682 branch and Zookeeper - installed via brew) and its working🚀. Ran a basic insert command

Attaching the HS2 logs and tez_am docker logs

tez-am-working.log working_hs2.log

Few things to note:

Add the following in /etc/hosts/ file. this is necesary for docker to communicate with hdfs running on localhost
127.0.0.1 host.docker.internal

tez-am docker image needs hive-exec jar i.e resorce localization. For now I added it using the plugin directory way i.e. -v "/Users/raghav/Desktop/plugin:/opt/tez/plugins" \ and plugin directory contains only hive-exec jar otherwise .
Vertex vertex_1769280834537_0000_1_01 [Reducer 2] killed/failed due to: INIT_FAILURE] ....
vertex=vertex_1769280834537_0000_1_00 [Map 1], org.apache.tez.dag.api .TezUncheckedException: java. lang.RuntimeException: java.lang.RuntimeException: java.lang.ClassNotFoundException: Class org.apache.ha
doop.hive.ql.io. CombineHiveInputFormat not found
core-site.xml
    <property>
        <name>fs.defaultFS</name>
        <value>hdfs://host.docker.internal:9000</value>
    </property>
hdfs-site.xml Connecting docker AM to datanode was really problematic, took lot of time 😅
    <property>
      <name>dfs.datanode.use.datanode.hostname</name>
      <value>true</value>
    </property>

    <property>
      <name>dfs.datanode.address</name>
      <value>host.docker.internal:9866</value>
    </property>

    <property>
      <name>dfs.datanode.hostname</name>
      <value>host.docker.internal</value>
    </property>
In docker tez-am tez-site.xml (Will update the PR for this as well in TEZ-4682)
    <property>
        <name>dfs.client.use.datanode.hostname</name>
        <value>true</value>
    </property>
Some --add-opens in tez-entrypoint.sh needs to be updated. Basically hive project --add-opens are also required. Will update the PR for TEZ-4682.

hive-site.xml
  <property>
      <name>hive.server2.use.external.sessions</name>
      <value>true</value>
  </property>

  <property>
    <name>hive.server2.tez.external.sessions.namespace</name>
    <value>/tez-external-sessions</value>
  </property>

  <property>
      <name>hive.server2.tez.external.sessions.registry.class</name>
      <value>org.apache.hadoop.hive.ql.exec.tez.ZookeeperExternalSessionsRegistryClient</value>
  </property>

  <property>
    <name>hive.zookeeper.quorum</name>
    <value>localhost:2181</value>
  </property>
NOTE/IMPORTANT: There was a flaky behaviour observed with tez-conf.pb in tez-staging directory in hdfs. It was throwing error for No file found. Not sure why it was occuring but it was fixed after some time automatically and not facing it again. Will post here if I face it again.

@Aggarwal-Raghav : this is awesome, thanks for testing! let me share the action items I can think of here to make this and TEZ-4682 happen:

/etc/hosts workaround: external setup steps, I wish we could get rid of them, but maybe it's crucial for the first iteration: I believe hive docker page can iterate about this: https://hive.apache.org/docs/latest/admin/setting-up-hive-with-docker/
tez-plugins folder: that's awesome, properly documented on TEZ-4682 side (if it's not enough, we can fully make it work out of the box in the scope of HIVE-29419

5/6: core-site.xml + hdfs-site.changes: these are must-have items on the hive site, however it would be better if I can have the hdfs setup and these xml configs at the same place, otherwise, the values might seem a bit vulnerable in a sense that no one knows under what circumstances host.docker.internal:9866 and hdfs://host.docker.internal:9000 are valid and working, this is what might be addressed by HIVE-29493

hive-related add-opens to tez config: in general, tez should not contain hive related stuff given that tez doesn't depend on hive, however, in case of add-opens args, I feel it might be beneficial to have them there as long as those don't refer to anything "hive", only java packages
hive-site.xml config: definitely something to be included here in this PR or a related one

abstractdog · 2026-03-09T13:40:05Z

@ayushtkn : I'm kindly asking you to review this huge thing :)
please consider @Aggarwal-Raghav 's manual testing as described here #6343 (comment)
basically Hive was able to submit a DAG to an external AM (given that zookeeper and hdfs was running on that machine), which was running by the image create in the scope of TEZ-4682

…or unmanaged, external Tez sessions

…static INSTANCES

sonarqubecloud · 2026-03-09T15:15:34Z

Quality Gate passed

Issues
56 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.3% Duplication on New Code

See analysis details on SonarQube Cloud

asf-ci-hive added tests pending tests passed and removed tests pending labels Mar 2, 2026

abstractdog force-pushed the HIVE-29477 branch from 66f4b84 to d9d3b2f Compare March 3, 2026 09:33

asf-ci-hive added tests pending tests unstable and removed tests passed tests pending labels Mar 3, 2026

abstractdog force-pushed the HIVE-29477 branch from d9d3b2f to 9afa704 Compare March 3, 2026 14:18

asf-ci-hive added tests pending tests unstable and removed tests unstable tests pending labels Mar 3, 2026

Aggarwal-Raghav reviewed Mar 3, 2026

View reviewed changes

abstractdog force-pushed the HIVE-29477 branch from 9afa704 to 9f48a61 Compare March 5, 2026 09:15

asf-ci-hive added tests pending tests passed and removed tests unstable tests pending labels Mar 5, 2026

abstractdog force-pushed the HIVE-29477 branch from 9f48a61 to 8e01971 Compare March 6, 2026 12:19

asf-ci-hive added tests pending tests passed and removed tests passed tests pending labels Mar 6, 2026

abstractdog requested a review from ayushtkn March 9, 2026 13:42

abstractdog added 5 commits March 9, 2026 14:58

HIVE-29477: Introduce TezSession interface and a new implementation f…

284f956

…or unmanaged, external Tez sessions

CuratorCache

f9ee488

removed static registry reference from different classes and rely on …

bd47c2d

…static INSTANCES

TestZookeeperExternalSessionsRegistryClient

738a65b

code cleanup

b3c0712

abstractdog force-pushed the HIVE-29477 branch from 8e01971 to b3c0712 Compare March 9, 2026 13:58

asf-ci-hive added tests pending and removed tests passed labels Mar 9, 2026

asf-ci-hive added tests passed and removed tests pending labels Mar 9, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

HIVE-29477: Introduce TezSession interface and a new implementation for unmanaged, external Tez sessions#6343

HIVE-29477: Introduce TezSession interface and a new implementation for unmanaged, external Tez sessions#6343
abstractdog wants to merge 5 commits intoapache:masterfrom
abstractdog:HIVE-29477

abstractdog commented Mar 2, 2026 •

edited

Loading

Uh oh!

Aggarwal-Raghav Mar 3, 2026

Uh oh!

abstractdog Mar 6, 2026

Uh oh!

Aggarwal-Raghav Mar 3, 2026

Uh oh!

abstractdog commented Mar 6, 2026

Uh oh!

Aggarwal-Raghav commented Mar 7, 2026 •

edited

Loading

Uh oh!

abstractdog commented Mar 9, 2026

Uh oh!

abstractdog commented Mar 9, 2026

Uh oh!

sonarqubecloud bot commented Mar 9, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

abstractdog commented Mar 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What changes were proposed in this pull request?

Why are the changes needed?

Does this PR introduce any user-facing change?

How was this patch tested?

Uh oh!

Aggarwal-Raghav Mar 3, 2026

Choose a reason for hiding this comment

Uh oh!

abstractdog Mar 6, 2026

Choose a reason for hiding this comment

Uh oh!

Aggarwal-Raghav Mar 3, 2026

Choose a reason for hiding this comment

Uh oh!

abstractdog commented Mar 6, 2026

Uh oh!

Aggarwal-Raghav commented Mar 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

abstractdog commented Mar 9, 2026

Uh oh!

abstractdog commented Mar 9, 2026

Uh oh!

sonarqubecloud bot commented Mar 9, 2026

Quality Gate passed

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

abstractdog commented Mar 2, 2026 •

edited

Loading

Aggarwal-Raghav commented Mar 7, 2026 •

edited

Loading