PHOENIX-7790 Add create command in PhoenixHAAdminTool for creating a new HAGroup

[ritegarg]

Summary

Adds a new create command to PhoenixHAAdminTool that allows administrators to create a new HA group entry in the SYSTEM.HA_GROUP table and initialize the corresponding ZooKeeper znode. This is a prerequisite step for setting up a new High Availability (HA) cluster pair.

Previously, HA group entries had to be created through lower-level programmatic APIs or test utilities. This PR exposes a CLI-based workflow consistent with the existing admin tool commands (update, get, list, initiate-failover, abort-failover).

What Changed

PhoenixHAAdminTool.java (+259 lines)

• New create command: Registered as CMD_CREATE in the main run() dispatch, alongside existing commands.
• executeCreate() method: Parses CLI arguments for both clusters using a symmetric slot-based model (cluster 1 / cluster 2 instead of local / peer), checks for existing entries (idempotent), inserts into SYSTEM.HA_GROUP via insertIntoSystemTable(), and triggers ZK znode initialization via HAGroupStoreManager.getHAGroupStoreRecord().
• haGroupExistsInSystemTable(): Queries SYSTEM.HA_GROUP using ConnectionUtil.getInputConnection() to check for pre-existing entries, enabling idempotent behavior.
• insertIntoSystemTable(): Performs an UPSERT INTO SYSTEM.HA_GROUP with all slot-based columns (ZK_URL_1/2, CLUSTER_URL_1/2, CLUSTER_ROLE_1/2, HDFS_URL_1/2, POLICY, VERSION).
• parseClusterRole(): Validates and converts string role values to ClusterRole enum.
• 8 new CLI options: --zk-url-1, --cluster-url-1, --cluster-role-1, --hdfs-url-1 and their cluster-2 counterparts.
• Help text and usage examples: Updated printUsage() and added printCreateHelp().
• Supports --dry-run to preview what would be created without persisting changes.
• Supports optional --admin-version (defaults to 1).

PhoenixHAAdminToolIT.java (+144 lines)

• testCreateCommandNewHAGroup: Verifies end-to-end creation - checks SYSTEM.HA_GROUP row is written with correct columns and ZK znode is created with matching fields (policy, cluster URLs, HDFS URLs, admin version).
• testCreateCommandAlreadyExists: Verifies idempotency - running create on an existing group prints a skip message and does not overwrite the existing row.
• testCreateCommandDryRun: Verifies dry-run mode shows planned creation without writing to SYSTEM.HA_GROUP.
• testCreateCommandMissingRequiredArg: Verifies RET_ARGUMENT_ERROR is returned when a required option (e.g., --policy) is omitted.

Design Decisions

• Symmetric slot-based options: The create command uses --zk-url-1/2, --cluster-url-1/2, etc., rather than --cluster-url / --peer-cluster-url (which the update command uses). This avoids the local/peer distinction and lets the same command run identically on both clusters.
• Idempotent: If the HA group already exists, the command returns RET_SUCCESS with a skip message, making it safe to re-run.
• ZK znode initialization: After inserting the system table row, the command calls HAGroupStoreManager.getHAGroupStoreRecord() which lazily initializes the ZK znode from the system table entry.
• Connection via ConnectionUtil.getInputConnection(): Used instead of manually constructing JDBC URLs, per project convention.

Usage

phoenix-consistentha-admin-tool create \
  -g myHAGroup -p FAILOVER \
  -zk1 zk-host1:2181:/hbase -c1 hmaster1:16000 -cr1 ACTIVE \
  -zk2 zk-host2:2181:/hbase -c2 hmaster2:16000 -cr2 STANDBY \
  -hdfs1 hdfs://host1:8020 -hdfs2 hdfs://host2:8020

Run this same command on both clusters to set up the HA pair.

Test Plan

• testCreateCommandNewHAGroup - verifies SYSTEM.HA_GROUP row and ZK znode creation
• testCreateCommandAlreadyExists - verifies idempotent skip behavior
• testCreateCommandDryRun - verifies no side effects in dry-run mode
• testCreateCommandMissingRequiredArg - verifies argument validation

[tkhurana]

Better to use ConnectionUtil.getInputConnection() than constructing the url by hand.

[tkhurana]

Same as above

[tkhurana]

Is this comment correct ? It seems the create command is also creating the znode.

[ritegarg]

Yes it is, fixed

[tkhurana]

Can we expose this method as an API also ? It will be useful from a test perspective.

[ritegarg]

Discussed offline and we plan to pick this later when time permits.