Review highlighted Cloudera Migration Assistant issues resolved in 4.0.0 for this release.
Cloudera Migration Assistant 4.0.0
- AMCM-3563: appendLogContents throws lock timeout exception at large HDFS scan
- 4.0.0
During very large HDFS scans, the server called appendLogContents faster than work could finish, which triggered database lock timeouts. The change paces that work so log upload no longer contends on the same lock during heavy scans.
- AMCM-3836: HDFS scan performance issue when FS_LOCATION table is full
- 4.0.0
When the FS_LOCATION table grew very large, HDFS scans slowed sharply because the server did expensive work against that table. The fix reduces that cost so scan progress stays usable on busy clusters.
- AMCM-3844: Hdfs items in plan are loading slowly with numerous paths in other plans
- 4.0.0
Opening the tree for an HDFS source in one plan could spend seconds per level if other plans contained thousands of stored paths. The UI query path was tightened so tree navigation stays responsive even when many paths exist elsewhere.
- AMCM-3825: Lock remains on the FS_LOCATION table
- 4.0.0
Concurrent updates against FS_LOCATION (for example clearing snapshot flags) could wait until the statement timed out while locking FS_LOCATION. Lock handling and transaction scope were adjusted so those updates complete reliably.
- AMCM-3433: Hive Scanning with Kerberos failure
- 4.0.0
Hive scans failed under Kerberos with JDBC handshake errors (GSS initiate failed) because the driver URI and principal wiring did not match how the CMA Agent authenticates. The agent now builds the Hive JDBC connection so Kerberos negotiation succeeds.
- AMCM-3583: HDFS2HDFS B2B Migration fails at Cloudera Replication Manager policy submit with Missing destination path
- 4.0.0
Base-to-base HDFS migrations could fail when submitting the Cloudera Replication Manager policy with HTTP 400 and a missing destination path in the create-policy response. Policy generation now supplies the destination layout Cloudera Replication Manager expects so submission succeeds.
- AMCM-3367: SQL Workload Scan Fails with in M-A Mode unable to SSH
- 4.0.0
In Master-Agent mode, SQL workload preparation still tried to SSH from the CMA Agent host even when SSH was not available or permitted. The flow was corrected to use in-process and supported execution paths so you are not blocked by SSH reachability for those scans.
- AMCM-3376: .sql is not Migrated at SQL Workload Migration
- 4.0.0
SQL workload migration picked up table metadata but left the selected .sql artifact behind on the source, so the migration dashboard could not show the file as moved or ready to remap. The pipeline now treats the script like other workload assets so it appears on the target side for review and mapping.
- AMCM-3391: --vault arg not working with distributed mode and ssl
- 4.0.0
Vault integration with TLS in distributed deployments did not propagate SSL settings correctly when Vault was enabled, so Vault looked active while SSL stayed off. Argument parsing, environment handoff, and tests were fixed so Vault and SSL settings reflect the configured deployment mode.
- AMCM-3452: Ozone scan can't be canceled when the ozone server isn't responding
- 4.0.0
Java Ozone scans could hang retrying a dead or overloaded Ozone endpoint, and cancel commands (UI or /api/clusters/{clusterId}/commands/cancel) did not unwind the work. Cancellation now tears down the stuck client path and related sync so you can stop a runaway scan without waiting for a long RPC timeout.
- AMCM-3552: No way to change HDFS Java scan user
- 4.0.0
HDFS Java scans always ran as the Cloudera Migration Assistant Agent service account. You can now configure which Cloudera Manager service's process user or Kerberos principal is used for scans by setting the CMA_MIGRATE_WITH_USER_OF_SERVICE environment variable (for example, HDFS to scan as the HDFS service user). This does not allow specifying an arbitrary Unix username.
- AMCM-3694: Scan doesn't delete children if no child remained in the parent
- 4.0.0
After incremental scans removed the last child under a parent, stale child rows could remain in Cloudera Migration Assistant metadata. Cleanup now removes those orphans so the catalog matches what is actually on the cluster.
- AMCM-4111: Access token may exceed header size limits on clusters with many groups
- 4.0.0
Access tokens embedded full role and policy objects, producing oversized Authorization headers and HTTP 431 errors on clusters with many groups. Tokens now carry role names and a version counter; servers resolve full roles through a versioned role catalogue cache.
- AMCM-4114: Browser column selector icon not fully visible in some layouts
- 4.0.0
The column selector icon in data browsers could be clipped or partially hidden in some layout configurations. Layout and icon placement were adjusted so the control remains fully visible.
- AMCM-4119: Ozone/HDFS scan type selection UI element is misaligned
- 4.0.0
The scan type selector for Ozone and HDFS scans was misaligned in the scan configuration UI. Alignment and spacing were corrected.
- AMCM-4131: Execution Step Settings drawer layout is broken
- 4.0.0
Command text in the Execution Step Settings drawer could overflow and break the drawer layout. Overflow handling and layout constraints were fixed so long command text displays correctly.
- AMCM-4133: Execution page Review Settings panel fails to load
- 4.0.0
The Review Settings panel on the Execution page could fail to load or render. The underlying API and UI wiring were corrected so the panel loads reliably.
- AMCM-4147: Migration Plan title field cannot be cleared once set
- 4.0.0
After setting a migration plan title, you could not clear the field to remove the title. The form now allows clearing the title value.
- AMCM-4162: Discovery page triggers duplicate services API call when selecting a collection
- 4.0.0
Selecting a collection on the Discovery page triggered duplicate services API requests. The services list query was deduplicated so selection performs a single fetch.
- AMCM-4170: Triggered By field overflows its container in the Cloudera Data Engineering info panel
- 4.0.0
In the Cloudera Data Engineering info panel, the Triggered By field could overflow its fixed-height container. Layout constraints were adjusted so long values wrap or truncate correctly.
- AMCM-4209: Delete Copied Item button non-functional on Git target side in Cloudera Data Engineering Airflow Metadata tab
- 4.0.0
On the Git target side of the Cloudera Data Engineering Airflow Metadata tab, Delete Copied Item did not remove mapped entries. The delete action now calls the correct endpoint and updates the UI.
- AMCM-4220: Commands list does not refresh automatically when new commands start while the page is open
- 4.0.0
The Commands list did not update when new commands started while you kept the page open. The UI now refreshes the list when new command activity is detected.
- AMCM-4222: Cloudera Replication Manager App credential suggestion shows masked password value instead of the usable credential
- 4.0.0
Credential suggestions for the Cloudera Replication Manager App showed a masked password placeholder instead of the value you could use. Suggestions now surface the usable credential reference.
- AMCM-4229: HDFS to Ozone same-cluster migration fails
- 4.0.0
Same-cluster HDFS-to-Ozone migrations could generate Hive ALTER TABLE … SET LOCATION SQL that still pointed at HDFS paths after files were moved to Ozone. Generated SQL now uses the correct Ozone locations, including for partitioned tables.