Kinetica Release Notes

Build version:

  •  350 Resolved Issues


KDEVOPS-497 Fix gpudb home owned by root
KDEVOPS-854 Adjusted maven central repo URL in thrift to https
KDEVOPS-854 Fix type-o in
KDEVOPS-854 Update thrift to 0.13 to fix http/https issue
KDEVOPS-863 Fix rank IP calculations in
KDEVOPS-866 Move creation of home dir out of %install in RPM
KECO-1375 Renamed destination tables to fit Kinetica table name and length limit
KECO-1375 Upped the version and updated changelog
KECO-1375 Wrapped native pyarrow.parquet.ValueError error in odo.KIOValueError
KECO-1477 Added column name normalization for CSV header
KECO-1477 Spark Connector version update
KECO-1542 Adding another lowercase inclusion check
KECO-1542 Updating test harness to handle key case
KECO-1620 Restoring original test expectations
KECO-1683 Added batch processing to esri files
KECO-1687 Fixed Java API dependency to exclude the 7.1 snapshot
KECO-1697 Fixed the exist() call for indirect case sensitive sources
KECO-1697 Removed quotes from comparison values in unit tests
KECO-1697 Treating source case sensitivity as deprecated flag
KECO-692 Verified the fix and updated test case outcome
KGEO-717 Added the 'aggregated_output' option to the MSD solver of /match/graph for unpivoting the output records per truck-store id route scheduling
KGEO-728 Added the create/modify graph options for setting the angle for turns turn_angle (default is 60 degrees)
KGEO-728 Incorporated creating of turns into create and modify graphs as a new option 'add_turns' (default=false) and respective left, right, intersection and sharp_turn penalties to the options of solve and match graph endpoints (defalut=0.0). Also added the test_dc_turn_penalty example in alltests
KGEO-735 Add Class-break track rendering
KGEO-735 Build fix for Attributes.h. std::vector::insert(iter) does NOT return an iter in gcc 4.8.5, change to void plot2d::PlotAttributes::append()
KGEO-735 CairoRenderer: fix aliasing warnings
KGEO-735 TrackProcessor: fix warnings
KGEO-737 Added a fast path for linestring/polygon st_intersection (also multipolygons with 1 polygon). -st_intersection now has an optional 3rd 'fast_path' parameter which defaults to 0 (off). If set to 1 then it will use the optimized path for linestring/poly intersection. -st_intersects fast path handles linestring/multipolygon when the multipolygon has 1 polygon component
KGEO-737 Improvements for geojoins, particularly st_intersects/st_intersection between a linestring and a polygon. -R-tree uses std::vector instead of std::unordered_set for collecting probe results (vector needs to be sorted/uniqued to remove duplicates, but testing showed this was still faster than the set approach. -Conversion from GEOS to Boost uses emplace_back (minimal speedup)
KGEO-740 Disable nv advanced blending
KGEO-740 Use glFinish and in more places
KGEO-740 Use glMemoryBarrierByRegion before FBO read
KGEO-742 Added new 'st_linestringfromorderedpoints3d' geospatial aggregate function
KGEO-742 Fix st_linestringfromorderedpoints3d sinking
KGEO-748 Make validation of WMS string parameters case-insensitive
KGEO-749 Refactorized and reimplemented portions of the Heuristic Travelling Salesman solver with parallel pairwise dijkstra runs so that at scale performance is satisfied added two new options to satisfy the SLA requirements for number of best paths from each location and the cap on the maximum number of combinatorial sequences
KINFRA-1275 Support switchover of hostmanager to another node Basic support via /admin/switchover endpoint Only moves ranks, no other head node processes
KINFRA-1276 Add missing call to close socket
KINFRA-1276 Change the SocketMonitor to more generically use ProcessId than rank number
KINFRA-1276 Change to always send comm failure event from ranks to local HM, which then forwards to head HM
KINFRA-1276 Detect rank connection broken using a ZMQ socket monitor
KINFRA-1276 Failure detection of one or more rank crash and trigger failover
KINFRA-1276 Fix problem in socket monitor class and move into its own file
KINFRA-1276 Fix rank hanging during shutdown if another rank restarts
KINFRA-1276 Fix some migration logic during network failure
KINFRA-1276 Set the rank statuses to not responding when the zmq connection is lost
KINFRA-1277 Fix host role status in debug endpoint Minor refactoring for avro name consistency
KINFRA-1277 Host leader election during switchover Core raft algorithm impl, only supports a directed switchover Fixed several bugs related to syncing GR values with GM after host migration TODO: move to its own class in separate thread outside GR TODO: use direct peer communication between hosts when head is inaccessible
KINFRA-1277 Host switchover & leader election improvements Detect multiple invalid host status scenarios (multiple leaders, inconsistent terms, etc.) and start new election Check connection to GM before starting leader election Demote hosts that are unreachable Notify ranks of new GM Notify ProcessMgr, MigrationMgr & Exe groups of new head node Trigger leader election on lost connection to head Migration doesn't currently work when other ranks besides rank-0 are on the head node (regardless of rank or host switchover/failover)
KINFRA-1277 Move leader election processing to dedicated class Runs in separate thread on HostManager Use direct peer communication between hosts via req/rep Supports host election via switchover or leader abdication (failover)
KINFRA-1359 Fix some migration bugs related to successive failures
KINFRA-1361 Add ability to switchover ranks from HM
KINFRA-1361 Add host sanity checks for migration
KINFRA-1361 Add initial commit of switchover implementation
KINFRA-1361 Fix some state bugs during switch/failover
KINFRA-1384 Add support for passive nodes
KINFRA-1384 Add support for passive ranks to test harness
KINFRA-1384 Update rank0_ip_addr in conf when migrating rank0
KINFRA-1389 Add migration data loading policy Same options as load_vectors_on_start
KINFRA-1393 Allow an orphaned node to resync the leader status after it has reconnected to the cluster Block leader election (and subsequent migration) when a quorom cannot be established from reachable hosts Fix bug in determining next leader Fix a hang when trying to restart the GlobalRegistry thread
KINFRA-1393 Compilation fix gcc 4.8.2 (std::vector::erase() only supports non-const iter)
KINFRA-1393 Fix some leader election bugs where head node migration fails if current head is unreachable LeaderElectionMon triggers election when head is unreachable and system idle (may be temporary) Skip migration of disabled processes Removed ODBC server status and type from ProcessType enum (this has been supplanted by queryplanner) Ranks can now be gracefully shutdown after Rank-0 has been migrated
KINFRA-1393 Leader election verifies that the candidate that requested votes is in passive hosts list
KINFRA-1396 Change INFO log to DEBUG on clear_table response
KINFRA-1396 Clear temp tables after a rank migration
KINFRA-1396 Suppress 'Table doesn't exist' errors when clearing temp tables
KINFRA-1400 Remove 'HostManager' process type and host migration logic from MigrationManager Fix couple bugs that prevented switchover endpoint from being run on non-head hostmanager
KINFRA-1434 /admin/remove/host now waits for host to go back to gaia mode/status of standby/stopped after removal
KINFRA-1434 Add ability to remove a host from the cluster using /admin/alter/host
KINFRA-1434 Break out /admin/alter/host into three separate endpoints
KINFRA-1434 Bump up gaia info version when host is added to or removed from cluster
KINFRA-1434 First cut of endpoint /admin/alter/host
KINFRA-1434 Fix issue where loop was holding scoped lock longer than is apparent
KINFRA-1434 Fixed issue where host removed from cluster couldn't be re-added, if it used to be a leader
KINFRA-1434 Updates to /admin/alter/host
KINFRA-1435 Add head/worker_failover conf params
KINFRA-1435 Add logic to handle np1 conf params
KINFRA-1435 Fix missing np1 header for conf params
KINFRA-1437 Add initial failover support to MigrationMgr
KINFRA-1437 Add initial poc impl for MigrationManager
KINFRA-1437 Add ProcessManager
KINFRA-1437 Update GlMgr to remove failed migrations
KINFRA-1437 Updated MigrationManager
KINFRA-1458 Add initial impl of calcite migration as part of rank0 migration
KINFRA-1467 Fix query planner not migrating
KINFRA-1467 Update host component table on debug page
KINFRA-1468 Add failover logic for the query planner
KINFRA-1470 Add failover trigger from httpd failure
KINFRA-1478 HA in core initial implementation: details in feature/ha
KINFRA-1484 Add additional logic for handling multi host network failure
KINFRA-1485 Add initial logic to handle network failures
KINFRA-1486 Add common function to get migration hosts
KINFRA-1489 Update leader selection logic to account for both passive & degraded hosts Suppress leader election requests when any cluster operation is in progress Remove test code and replace 'host' with 'head' migration in switchover endpoint for clarity Expose current leader in gaia_info struct Additional error checks and debugging msgs
KINFRA-1489 Use the cluster_leaader field in gaia_info struct to trigger leader change notification Restore destination host head switchover capability
KINFRA-1499 Add support for show system properties to Worker ranks
KINFRA-1512 Added HA conf params to replace current xml ones (in HA proc).  * This will be eventually be used with etcd + Fixed an issue with the shutdown and recovery when Rabbit is not responsive
KINFRA-1512 HA gpudb conf changes:  * Added ring and cluster name  * Removed HA conf from gpudb.conf with the exception of enable_ha  * Using a tmp json file for HA conf * Update show_system to still send the old HA format this should be switched out
KINFRA-1520 Fix startup issue
KINFRA-1520 Handle couple of errors when rank 0 is migrating simultaneously with another rank
KINFRA-1524 Add leader election prior to rank0 group failover
KINFRA-1524 Update leader election to stop migrating workers
KINFRA-1531 Minor refactor in GM
KINFRA-1531 Update election debug stmt
KINFRA-1531 Update GM host heartbeat logic
KINFRA-1531 Update httpd conf params during migration
KINFRA-1541 Updated processes to use restart attempts from conf
KINFRA-1547 Add conf event_server_address and port, update resourcegroup metrics, basic event for sql queries
KINFRA-1547 Add Monitoring/Metrics and SystemEvent, StatsLogger is not used
KINFRA-1558 Fix several migration bugs
KINFRA-1558 Fix some more bugs during rollback testing
KINFRA-1558 Fix various failures discovered with the test harness
KINFRA-1558 Update /admin/switchover inputs
KINFRA-1558 Update test harness to add test to force switchover failure
KINFRA-1559 Expose host roles and leader status in /show/system/status
KINFRA-1560 Refactor test_agent_harness into dart library
KINFRA-1562 Add failover support for graph initial commit
KINFRA-1563 Add initial text/multiprocess linkage commit
KINFRA-1563 Add second flag for GM messages
KINFRA-1563 Avoid clobbering GR MIC version
KINFRA-1563 Fix errant text server restart flag clear
KINFRA-1563 Temporary fix for updating MIC during head change
KINFRA-1563 Update migration retry logic to use session id
KINFRA-1570 Fix rank0 shutting down cluster after failover
KINFRA-1571 Added guard to prevent leader abdication when no hosts are available for failover Added means to request role_info of any/all hostmanagers from a rank Avoid checking vote tally for election terms already finalized (elected w/o declaring candidacy error) Fixed bug that lead to heartbeat processing and leader election being blocked on a host when all workers/processes are migrated off to another host Fixed a race condition that lead to an election restarting because the status was reverted from candidate to follower Addded post-election cluster verifiction check to compare each node's role_info status
KINFRA-1571 Dont begin failover to same host
KINFRA-1572 Add thread pool to GM
KINFRA-1574 Add rank restart to conf
KINFRA-1574 Add rank_restart_limit to show sys props
KINFRA-1579 Fix 'Host not found' errors generated on startup and occasionally during host migration
KINFRA-1579 Sync role_cluster_info after HM restarts or host gets out of sync Fix bug that lead to election timing out due to multiple threads waiting on same resources Fixed a LE regression bug caused by GM/GR on new head being out of sync with each other Suppress excessive messages in LE startup while waiting for cluster status to sync Suspend heartbeat message while cluster op in progress or the leader isn't set Log the head GM ip sending heartbeats to aid with debugging Move heartbeat handling to its own function
KINFRA-1589 Fix various issues with version mismatches with heartbeating enabled
KINFRA-1589 Re-fix inf waiting for logger on debug page
KINFRA-1589 Refactor rank 'not_responding' into cluster.rank_info
KINFRA-1589 Update GR to not use infinite timeout for send_start_count
KINFRA-1589 Update head GM to do heartbeating
KINFRA-1590 Update ProcessMgr to route rankInfoClst data to exeProcs
KINFRA-1591 Add cloud_disk_info.json
KINFRA-1591 Fix script not unmounting drives if there are multiple mounts
KINFRA-1591 Fixes for mount script interaction
KINFRA-1591 Handle cloud_disk_info_cluster msg in GR
KINFRA-1591 Initial check-in for supporting Azure deployment with dedicated disk volumes for persist data
KINFRA-1591 Update MountMgr error reporting to include rc
KINFRA-1591 Update rank sync timeout to 2 minutes
KINFRA-1594 Add support for rank0 to use Persist tier folder
KINFRA-1594 Added more checks for various cases
KINFRA-1594 Create a separate mount manager to handle attaching and detaching volumes
KINFRA-1594 Fix broken test, silence warning re unused variable
KINFRA-1594 Fix default test cases in harness
KINFRA-1594 Fix EPDebug.cpp compile error
KINFRA-1594 Fix issue with detach aborting when uuid is specified
KINFRA-1594 Fix issue with missing data volumes sharing common mount point
KINFRA-1594 Improve mount data volumes handling in azure
KINFRA-1594 Refactor DataVolumeMountManager to MountManager
KINFRA-1594 Refactor mount detection logic from GR to MountMgr
KINFRA-1594 Update dart framework to support public ips
KINFRA-1594 Update getMigrationGroup logic to account for volumes
KINFRA-1637 Add host_number (hostN) to the hosts schema output in /show/system/status
KINFRA-1637 Host conf updates initial commit Add support for hostN entries where = hostN Remove np1 passive/degraded entries, instead use hostN.accepts_failover Added heartbeat_missed_limit option in conf Prevent /show/system/status from overwriting local GR's gaia_info struct Additional guards to prevent multiple hosts writing same test file when verifying a persist directory in a shared directory environment Let thru gaia_status updates on GM TODO: upgraded path for rankN.public_urls,rankN.taskcalc_gpus & numa
KINFRA-1641 Update worker_failover to match host_failover conf type
KINFRA-1642 Add DAG to MigMgr and update MountMgr to coordinate mounts
KINFRA-1642 Add support for ram/gpu limits during migration
KINFRA-1642 Fix ping check in MigMgr
KINFRA-1642 Fix temp tables not getting cleared
KINFRA-1642 Refactor MigMgr code in preparation for updates
KINFRA-1642 Refactor MigrationManager round 2
KINFRA-1647 Add failover when host connection lost
KINFRA-1647 Fix enum map call
KINFRA-1648 Generate unique disk-write-test filename while validating directories for r/w access in conf Race condition for NP1
KINFRA-1656 Add null check on async_exit_thread
KINFRA-1656 Attempt to fix rank0 setting cluster to standby during np1
KINFRA-1656 Fix rank0 not available after rank0 failover
KINFRA-1657 Add validation to allow only IPC for sql planner and text server
KINFRA-1657 Change head/worker_failover config to a boolean
KINFRA-1657 Rename rank_restart_interval -> restart_interval
KINFRA-1661 Restore public_url support for NP1 No longer reading rank0_ip_address and public_urls from conf, they are now derived from the hostN entries Add rank.taskcalc_gpu list to hostN during upgrades Make NP1 params optional and use default values if key does not exist Reset GR heartbeat time when host_status changes to running, prevent delayed heartbeat warning when db offline Changed GR broadcast messages from INFO to DEBUG logs Added more comments to
KINFRA-1665 Head host failover support When the head is orphaned off, ranks are unreachable since their only link to the cluster is head GM Ranks now maintain a separate pub/sub connection to their local GM and can be notified of NP1 events GR::send_member_update will now make use of the update flag to mutate the object prior to send/recv Allow negative term numbers to indicate hosts that are unresponsive or in an invalid state Added more checks to prevent errant leader elections due to hosts being out of sync (network loss, restart, missed heartbeats, etc.) Restore code in LeaderElectionMonitor to trigger host comm failure when a host is orphaned (in lieu of hosts heartbeating each other) Auto-select the cluster leader for rank-0 destination after a head host failure Block worker switchover to a host that is unresponsive, ditto for head switchover Fixed race condition and deadlock in GR::reinitGR when restarting a previously failed host Fixed numerous bugs in leader election exposed by multiple host failures
KINFRA-1680 Fix rank timeout while mounting
KINFRA-1680 Fix stats thread inhibiting rank0 shutdown
KINFRA-1680 Update CommsUtils to never mark rank0 as inactive
KINFRA-1681 Regen conf in-place after 7.0 -> 7.1 conversion Fix upgrade bug, regression from prior merge
KINFRA-1684 Fix planner fails to migrate if next sess id is 32
KINFRA-1702 Fix 4+ rank migration deadlock
KINFRA-1702 Fix host not reporting degraded state
KINFRA-1702 Fix rank not marked as mounting after failover
KINFRA-1702 Fixes for some state transition race conditions during failover
KINFRA-1702 Update MigRq wrapper to disallow duplicate entries
KINFRA-1702 Update recreate tables log to include error
KINFRA-1710 Fix version check breaking shard info updates
KINFRA-1710 Initial updates/minor fixes related to this ticket
KINFRA-1710 Update migration checks to allow inf wait for mounting
KINFRA-1729 Fix segfault during zmq::sendMsgToRankTag
KINFRA-1730 Add minor logging improvements to cluster ops
KINFRA-1730 Add sync state to coordinate between GR and GM threads
KINFRA-1730 Add temporary debug logging to track down hb size issue
KINFRA-1730 Fix sync point logic
KINFRA-1732 Enforce Rank-0 migration to cluster leader after a failover/failback event Wait for Rank-0 to be in running state before clearing temp tables Added a short sleep to wait for HTTP server to come up Reset heartbeat timeout after leader change Additional debug logs in LEM
KINFRA-1740 Replace conf.gm_ip usage with gaia_info.cluster_leader Regenerated conf w/out rank-0 would lead to the param being invalidated and leading to migration failures Add a flag to skip host ping check after a peer comms failure Degrade previous head host after a failover to a new head
KINFRA-1740 Retry switchover after timeout Check that next expected host for switchover/failover is reachable Remove defunct GR code that resets member arrays in a request
KINFRA-1753 Another fix to get around hosts with differing conf on startup
KINFRA-1753 Extend GaiaZmq api to accept multiple addresses Fix for degraded head HM re-connecting to cluster establishing new GM Ips (works so long as head migration already complete) Minor stylistic changes in LEM class no functional changes Version log in SecurityManager for debugging
KINFRA-1753 First cut with support for head host rejoining the cluster after network failure Host is put in a 'degraded' state when it can no longer communicate with majority of cluster The rest of the cluster will proceed with leader election & migration provided they have quorum Ranks and processes are shutdown on the previous host if/when it rejoins the cluster Added a facility to broadcast GR members to all hosts and ranks through use of another pub socket in GM that all GRs subscribe to (need to add conf) Prevent socket monitor from sending comm failure for a rank that is migrating Block migrations from hosts that cannot establish quorum
KINFRA-1753 Fix compile error under 4.8
KINFRA-1753 Fix failed host migration errors from last commit Prioritise local hostmanager when requesting role_info_* Get contextual host information when pinging to prevent errant elections Reset orphaned/degraded flags after GR resets Use the SyncPoint to orchestrate the restart of all GR's before triggering migration post leader election
KINFRA-1757 Added the functionality of Watch. Fixed few review comments
KINFRA-1757 Initial implementation of ETCD API
KINFRA-1764 Clear temp tables after a rank restarts
KINFRA-1764 Fix gcc6 compile clusterOps
KINFRA-1764 Wait for all ranks to reach running state post migration before clearing temp tables, not just R0
KINFRA-1766 Fix exception in /show/system/properties when running on localhost only Destroy GlobalManager earlier to fix GlobalRegistry ref count error on shutdown
KINFRA-1766 Update worker urls in /show/system/properties when httpd enabled
KINFRA-1767 Add support for tier.persist.graph0
KINFRA-1767 Update MigMgr graph dependency for graph server
KINFRA-1772 Fix 'Index exceeds hb slots' error after np1 event
KINFRA-1772 Fix Index exceeds rank hb slot after np1 event
KINFRA-1786 Fix shared mount ranks not migrating
KINFRA-1792 Remove reliance on sudo, error checking and added help
KINFRA-1792 Revert back to using sudo for disk mounting
KINFRA-1801 Fix 'Persist Aborting read due to cancellation' after migration
KINFRA-1801 Fix build failure
KINFRA-1814 Fix add ranks on np1 branch
KINFRA-1814 Fix add ranks with text server
KINFRA-1814 Fix adding a new rank in place of a previously removed rank
KINFRA-1814 Fix remove ranks on np1 branch
KINFRA-1814 Fix remove ranks with text server enabled
KINFRA-1814 Fix
KINFRA-1814 Small fixes for add/remove ranks
KINFRA-1822 Add the storage Volume info to the gpudb.conf
KINFRA-1822 Fix compiler warning and other minor cleanups
KINFRA-1823 Validate storage volumes are mounted at startup when storage script is provided
KINFRA-1824 Fix to allow HM restart from standby
KINFRA-1825 Add GlobalManager local pub port to conf Deprecate rank.public_url, fully replaced by host.public_url_prefix Remove unsupported argument from rank command line args to start GlobalManager Fix rank startup failure if GM pub port configured different from default (gm_port1 + 1) Deprioritize non-responsive hosts when doing req/rep GR::sendGMPing now handles exceptions internally vice the caller having to wrap it in a try/catch Bail out of ZMQSocketMonitor event loop if GRSingleton is null (cluster going to standby)
KINFRA-1826 insert_records_from_payload_request
KINFRA-1826 Refactor ExternalFilesReader setColumnsToLoad
KINFRA-1826 Send correct error on failure, on insert from payload
KINFRA-1826 Update insert records from payload endpoint
KINFRA-1827 Insert records from files use insert from payload when in head loading mode
KINFRA-1835 Add check to verify migration was to a valid host
KINFRA-1835 Honor host's accepts_failover flag when failing over Rank-0
KINFRA-1851 Fix 'job is hung' errors after switchover
KINFRA-1851 Fix some bugs in the harness/fuzzer
KINFRA-1858 Enable /show/system/status on worker ranks
KINFRA-1858 Renamed hostN.ip_address -> hostN.address, hostN.public_url_prefix -> hostN.public_address Update /show/system/properties to show all available interfaces public urls per host Fix regression bug where all interfaces/urls by rank weren't properly generated by endpoint Add storage volumes to /show/system/properties
KINFRA-1860 Fix rank0 lack of responses to types broadcast
KINFRA-1860 Set socket option before socket connect in rank monitor
KINFRA-1864 Align host entries in host_info_cluster based on their order in the conf (hostN.*) Enforce unique host addresses in conf (internal, public & host_manager) Update conf.head_ip_address on head migration Added file version to conf, for now just the build version Revert GlobalRegistry member value if send_recv fails
KINFRA-1864 Update conf to specify to use 'host#' instead of just the host index Added more host and persist validation checks in conf Add graph persist path to /show/system/properties
KINFRA-1868 Call SetIndexedDb::initSetIndexedDb for sparse check in TypeObjectManager ctor
KINFRA-1870 Fix graph client not working after migration
KINFRA-1870 Fix migration stuck if it fails in the middle
KINFRA-1870 Fix table is None error in test harness InsTask
KINFRA-1870 Refactor ClusterOps
KINFRA-1870 Update rank mount processing
KINFRA-1878 ExeProcess now waitForPid(120) in thread, then kill -9, and abandon pid
KINFRA-1879 Flush/Fsync indexdb records_info file along with its object store and json metadata according to conf IndexedDb was reporting missing metadata files and would rescan the TOC to regenerate it each time
KINFRA-1881 Add etcd status publish from HostManager
KINFRA-1881 Update conf to remove default etcd host
KINFRA-1881 Update GM to recreate the etcd client whenever the conf changes
KINFRA-1886 Moved default HA modes for endpoints to etcd conf
KINFRA-1886 Moved HA conf to etcd + some etcd client cleanup
KINFRA-1889 Use unique inproc address everytime to prevent 'Address already in use' error when making new connection after rank restart
KINFRA-1895 Update ExeProcess to kill process if startExe fails
KINFRA-1898 Do not prevent tables created by system user from being added to SecurityManager
KINFRA-1901 Enable multihead by default in conf A warning is logged on startup if multihead is disabled but N+1 is enabled Added GaiaParams utility functions for retrieving host_conf Verify graph directory on startup Remove git datestamp from conf file version
KINFRA-1901 Possible fix for race condition when altering cluster configuration host_info_cluster was was being broadcasted before it had been fully resized to account for the new host The resize also depended on the conf being updated Add a syncpoint after expanding the conf just before bringing the host into the cluster
KINFRA-1901 Update packaged file_version = 7.1.0.... in gpubd.conf and gpudb_logger.conf in
KINFRA-1907 CsvFile change ManagedVector to ManagedString
KINFRA-1916 Account for non-rank tiers when resizing the tier params
KINFRA-1921 GaiaSetDeferredDataLoader needs its own TierExecutionContext
KINFRA-1930 SetManager populate_persisted_child_tables() erases orphaned children tables if found. Fixes segfault, cause was a throw during a table delete
KINFRA-1951 Prevent GaiaSetManager from syncing projections during construction or tables that have failed a rebuild
KINFRA-1956 Fix comms 'unblocking jobs' race condition
KINFRA-268 Add degraded hosts logic for process failure
KINFRA-268 Check in missing schema
KINFRA-268 Remove most of graph coupling to rank0 for migration purposes
KOLAP-1712 Add ScopedClearTable() to query execution. SqlDriver's global scopedcleartable tracks the all result tables that are created and clears them in the case of exceptions. The SqlPlanExecutor's local ScopedClearTable() tracks all the last_use work tables and clears them when the step execution is done
KOLAP-1712 Do not add existing tables to ScopedClear
KOLAP-1833 Test update. explain will show the result distribution for /aggregate/groupby endpoint in the SQL plan
KOLAP-1915 Fix for camel casing of SHOW SECURITY output
KOLAP-1915 Improve output of SHOW SECURITY (SQL Command)
KOLAP-1915 Sort output of SHOW SECURITY for consistent output, and ease of use to find things
KOLAP-1915 Update test to expect new SHOW SECURITY output format
KOLAP-1919 Avoid write lock on pkgen for workers when it is not necessary
KOLAP-1919 Cherry-pick the changes from 7.0. avoid write lock on pkgen when it not needed
KOLAP-1919 Ffixed conflicts due to cherry-pick merge resolution
KOLAP-1979 Support for /alter/table changing a column with Row/Column security applied to it
KOLAP-1979 When columns are renamed, update the column name in Row/Column security filter expressions
KOLAP-1981 Do not dictionary encode non-char-n string fields
KOLAP-1981 ExternalFileReader now inrfers type if header is not kinetica header. Python used to create external tables and reads all tables in all repos and trys to create the tables
KOLAP-1981 TypeExternalFileInfo stored in Type for easy debug access and access to date/time/datatime scanning formats. Enpoint_create_external_table respons info['external_file_info'] is set to a string in format that can be eval'd by python script and describes the FieldDescs and how the files was parsed.. The script now sucessfully reads most csv files found in kinetica repos and pretty-prints the FieldDescs
KOLAP-1983 Fix shadow cube tablekey type to be vector<string> instead of colon(:) seperated table names
KOLAP-1983 Rrevert the comment. and update a testcase to check that shadow cubes work with table names that have colon(:) in it
KOLAP-1986 Load persist table monitor properly and rebuild topic sets
KOLAP-1986 Persist complete table topic info instead of just topic_id
KOLAP-1986 Put back loadFromFile() function
KOLAP-1986 Save event type as string instead of int in table monitor json file
KOLAP-1986 Some necessary changes based on code review feedbacks
KOLAP-1986 Test if user can still get table monitor notification when table is renamed
KOLAP-1986 Unregister topic needs to update table monitor json file
KOLAP-1986 Update persist json file correctly when drop tables
KOLAP-1986 Update table monitor json file correctly after renaming table
KOLAP-1987 Add more show/table tests after calling clear/table/monitor
KOLAP-1987 Put table monitor info in additional_info
KOLAP-1987 Show/table shows table monitor info
KOLAP-1991 Fix for getting the correct projection count for replicated tables when there are multiple TOMs per rank
KOLAP-2001 A DATETIME oddesy. Add support for 1 or 2 character MONTH and DAY strings. Improved days_in_month to avoid calculations in most cases. Improved timestamp_to_char to handle embedded quoted strings in a single pass and not need to use pt_text. Some other code cleanup
KOLAP-2001 Allow T and Z in DATETIME strings (e.g., "2020-01-19T12:34Z") -- this is a common format for UTC times. Update days_in_month for better CUDA code & CPU code
KOLAP-2001 Support non-literal format string in TO_CHAR. Update newly added tests to use GPUdbTable. KOLAP-2013 Support default 2nd parameter in OBFUSCATE function
KOLAP-2012 Interpret the source table row as per the schema. -Handled the column names with aliases in the projection. -Handled the filter expressions in the create projection. -Modified the internal schema for create_projection_finalize to carry the base column names to the worker
KOLAP-2014 Re-enable compound equijoin according to the config
KOLAP-2018 Add more restart tests to
KOLAP-2018 Note GaiaSets are updated when chunk is actually deleted attempt to avoid incremental refresh with stale chunk-ids
KOLAP-2027 Implement Lazy/On-demand attribute/chunk_skip indexes
KOLAP-2027 Remove local unused test code from
KOLAP-2029 Fix to support more string types with column restrictions
KOLAP-2031 Added test for OBFUSCATE
KOLAP-2032 Fix for accessing columns available in some rows, when there are other columns restricted to other rows
KOLAP-2032 Fix for not combining row security with column security
KOLAP-2033 Added a TO_CHAR dynamic formatting test
KOLAP-2033 Changed new codes to not interefere with Oracle codes. Reduced the number of copies of MONTHDAYS and DAYNAMES. Minor optimizaiton to days_in_month
KOLAP-2033 Use a lambda function to reduce repeated code
KOLAP-2035 All KERNEL_EXEC_INIT calls moved from Worker_visualize_image_{heatmap/contour}.cpp int Plot2D{Heatmap/Contour}
KOLAP-2036 Reverted changes that got rid of ai fast path for inequalities. Modified is_ai_rpn so FilterPlanner would not use unary-ai path for expressions with inequalities
KOLAP-2039 Updated test for more flexibility in testing TTL
KOLAP-2041 Tables with TTL are not going away even if they don't change
KOLAP-2041 Test to confirm table goes away after TTL (even if show_table is called)
KOLAP-2042 For track tables where the shard key isn't specified we automatically set TRACKID to the shard key but now we also update the type properties for TRACKID to reflect that. This should fix the issue with track tables on startup printing out the warning about setting the shard key to TRACKID
KOLAP-2045 /aggregate/minmax on datetime works like on timestamp and returns values in milliseconds since the epoch. Also /aggregate/histogram and /aggregate/statistics
KOLAP-2045 Add min/max support for DATE and TIME columns
KOLAP-2051 Fixed missing cuda-syncs
KOLAP-2056 Do not check security on base tables for DML and let the endpoint do it
KOLAP-2085 Replace calls in FilterPlanner::extract_equijoin_plans type->get_any_column_name_for_set_index(set_index) with string +std::to_string(set_index)
KOLAP-2105 Add additional parameters 'has_primary_key, primary_key_set_index' to JoinedSet::get_chunk_filter_key to create keys correctly for primary key join case and modify Worker_primary_key_join filter to supply the primary_key_index
KOLAP-2105 Add additional test to test_join_kolap2105.p
KOLAP-2105 Add test that shows shadow cube errors however have not yet duplicated segfault
KOLAP-2105 Make const variables in loop references
KOLAP-2105 Revert last change JoinedSet::get_chunk_filter_key changed to loop over plan_steps in reverse order
KOLAP-2108 Fix to run well with nosetests.  add it to
KOLAP-316 Pull out functions fprintEvent, fprintBeginEv.ents and fprintEndEvents from DumpTracer::dump_summary_info
KUI-1163 Support non-default GPUdb port for KiSQL
KUI-1165 Command injection through getLogs endpoint
KUI-1174 Lock down server file import to external_files_directory from gpudb.conf for KIO and Advanced CSV
KUI-1175 GAdmin WMS map filtering polygon opens popup on click
KUI-1177 UI doesn't pass --d-property-additions along with --gis-ext-enabled when import from postgresql
KUI-1178 Add KIO import/export command preview
KUI-1181 Table Info modal not handling new rank 0 resource statistics properly
KUI-1186 Ingest from PostGIS passes two additional flags to KIO when performed via the UI
KUI-1187 WKT LINESTING if has a z component ;point, the client side renderer fails
KUI-1193 Remove unnecessary '=' from generated KIO commands
KUI-1197 Restrict access to external_files_directory folder when user runs INSERT INTO ... FROM FILE queries with kiSQL
KUI-1199 Allow unselecting Cold/Disk Tiers for better user experience



None currently


Should you have any questions or concerns, please visit our support page, official documentation page or email us at

Was this article helpful?
0 out of 0 found this helpful
Have more questions? Submit a request


Article is closed for comments.