Kinetica provides a SQL interface through its JDBC driver, which runs on
Linux, Mac, & Windows. Clients that can’t use JDBC can connect using the
PostgreSQL Wire Protocol. For
details on the extent of SQL compliance, see the SQL section.
Connecting
KiSQL
KiSQL is a Kinetica JDBC-based SQL CLI client that is available on GitHub.
Details can be found on the Kinetica SQL (KiSQL) page.
JDBC Applications
Java applications which use JDBC can reference the Kinetica JDBC driver in the
Maven Central Repository as a pom.xml dependence:
When connecting to a Kinetica instance with one or more failover clusters, the
connection string should be updated to contain a comma-delimited list of
head node addresses for all clusters. The URL of the primary cluster can also
be specified.Add a URL parameter to the JDBC connection string to contain all
cluster head node addresses and add a PrimaryURL parameter
designating the cluster that should be tried first.
URL of Kinetica database server to connect to—this may be a comma-delimited list of multiple clusters in an HA configuration
CombinePrepareAndExecute
0
Whether or not to combine the JDBC prepare and execute phases, when appropriate. This can improve the performance of static queries (those that lack parameterized values).
* 0 = Don’t combine prepare and execute phases * 1 = Combine prepare and execute phases
This parameter should be set to 1 whenever possible. Some JDBC clients may return odd errors when this is active, however; and may need to be set to 0 to function properly. This parameter is not needed when using KiSQL .
UID
User ID for database connection
PWD
Password for database connection
If the password contains special characters, they must be encoded using URL-style encoding; e.g., encode a semicolon as %3B
OAuthToken
OAuth token used in OIDC-enabled environments; when using, leave UID & PWD blank
Timeout
0
Minutes to wait for a response from the database; 0 for no timeout
Schema
<user default>
Default schema to use within the session
ImpersonateUser
Impersonate another user for this session. Only available when connected as an admin.
SessionId
<new for eachconnection>
JDBC only. ID of an existing session to join, so this connection shares its session temp tables. Letters, digits and - only, as in a UUID.
PrimaryURL
URL of a cluster to always attempt to go to first (if URL is a list)
TokenNameClaim
Name of the claim token identifying the username; sub is the default claim token name used.
Server Override Parameters
The following are optional parameters that could be used to override the default
settings, which are set by the server. See the
Configuration Reference for more information
on configuring the SQL Engine server-side.The default value for each of these options is 0.
Whether or not to process operations that would require resharding of data to complete successfully
* 0 = Disable distributed operations, failing any queries requiring resharding of data to succeed * 1 = Enable distributed operations, resharding data to meet sharding requirements, if necessary
IgnoreExistingPk
Whether or not to suppress errors for record collisions when inserting into or updating a table with a primary key; if the specified table does not have a primary key or UpdateOnExistingPk is 1, then this parameter is ignored
* 0 = Treat as errors any inserted or updated records whose primary keys collide with those of existing records * 1 = Ignore inserted or updated records whose primary key values collide with those of existing records
LateMaterialization
Whether or not to avoid materializing intermediary result sets when processing a complex query
* 0 = Materialize intermediary result sets * 1 = Avoid materializing intermediary result sets
PagingTableTtl
TTL for the result set paging table, in minutes; other possible values:
* -1 = no TTL (temp sets will not expire) * 0 = expire immediately (at next clean-up interval)
ParallelExecution
Whether or not to execute independent portions of a query in parallel
* 0 = Execute each component of a query in series, regardless of intra-query dependencies * 1 = Enable parallel execution of independent portions of a query
PlanCache
Whether or not to cache query plans, as derived by the rule-based/cost-based optimizer
* 0 = Disable caching of query plans * 1 = Enable caching of query plans
ResultsCaching
Whether or not to cache result sets, which happens automatically for certain queries that have only materialized views as sources
* 0 = Disable caching of result sets * 1 = Enable caching of result sets
Whether or not to use scalar subquery optimizations, in which non-correlated single-value-returning subqueries will be evaluated first and then those result constants will be used in the remainder of the query processing
TTL for temporary result tables, in minutes. Other possible values:
* -2 = use server’s default TTL * -1 = no TTL (temp sets will not expire) * 0 = expire immediately (at next clean-up interval)
UpdateOnExistingPk
Whether or not to handle primary key collisions using an upsert scheme for inserts and overwrite scheme for updates when operating on a table with a primary key; if the specified table does not have a primary key, then this parameter is ignored
* 0 = Reject any INSERT or UPDATE that results in a primary key collision with an existing record * 1 = Use the inserted or updated record to replace an existing record with the same primary key when there is a primary key collision between the two
UseApproxCountDistinct
When issuing any COUNT (DISTINCT x) query, it will be computed using the faster but less accurate APPROX_COUNT_DISTINCT() aggregate function.
* 0 = Compute and return accurate distinct count statistics * 1 = Return faster but less accurate distinct count values
ValidateChange
When issuing an ALTER TABLE that changes column types/sizes, verify that the column data can be contained losslessly in the target column type/size. If 1, fail the request if the data will not fit; if 0, truncate the data to fit.
* 0 = Truncate any column data that won’t fit into the new type size on a column change request * 1 = Validate any column change and fail the request if column data is too large/long for the new type/size
Client Parameters
The following optional parameters may be used to change the JDBC client behavior.
Config Parameter
Default Value
Description
CombineResults
0
How to return the results of multiple DDL statements in a single Execute call. (JDBC only)
* 0 = Return individual result counts for each DDL statement * 1 = Combine result counts into a single result
DisableAutoDiscovery
0
Whether to disable querying the server for its configuration. (JDBC only)
Disabling auto-discovery will, in turn, disable both multi-head ingest & multi-head egress, as the client won’t have the list of URLs for the cluster nodes, required for distributed operations.
Disabling auto-discovery will also disable high availability failover, preventing attempts at performing operations on another cluster after a failure.
Auto-discovery may also be disabled by the driver itself after connecting, in which case it logs the event and falls back to head-node-only operation and multi-head operations become unavailable. This happens if the cluster is reachable at the URL supplied by the client, but not at the server-configured address advertised upon connection (for example, a container or NAT’d address), or if the initial discovery attempt fails for any other reason.
* 0 = Perform standard auto-discovery * 1 = Do not perform auto-discovery
DisableFailover
0
Whether to disable high availability failover, preventing attempts at performing operations on another cluster after a failure. (JDBC only)
* 0 = Do attempt to perform the operation on another cluster after a failure, if another cluster is configured * 1 = Do not attempt to perform the operation on another cluster after a failure
DisableMultiheadInsert
0
Whether to use distributed ingest for supported INSERT statements. (JDBC only)
* 0 = Use high-performance distributed ingest when possible, unless KI_HINT_SERVER_SIDE_INSERT is specified * 1 = Use normal INSERT processing
DisableSnappy
0
Whether to disable Snappy compression when transmitting data to the server.
* 0 = Compress data using Snappy * 1 = Don’t compress data using Snappy
* Permissive = If an error is encountered parsing a source record, attempt to insert as much of the valid fields from the record as possible. * Skip = If an error is encountered parsing a source record, skip the record. * Abort = If an error is encountered parsing a source record, stop the data load process. Primary key collisions are considered abortable errors in this mode.
FailbackPollInterval
5
Number of seconds to wait between fail-back service polls of the primary cluster when failed over.
FailoverOrder
Controls the order in which the cluster to fail over to is selected.
* RANDOM = Choose a random cluster to fail over to. * SEQUENTIAL = Use the order in which the clusters were defined by the user in the connection configuration, or if not defined, the order in which the server defines the clusters, to choose the cluster to fail over to.
FakeTransactions
0
Whether to throw exceptions when the client attempts to start or rollback a transaction; useful for 3rd party tools, such as JPA or Hibernate, that expect transaction support. (JDBC only)
* 0 = Throw an exception when client attempts to control a transaction * 1 = Do not throw an exception, but still ignore, if client attempts to control a transaction
Kinetica does not support transactions; this will only affect how the JDBC driver responds to transactional requests it can’t fulfill.
FileReadComment
#
When inserting data from a file with INSERT INTO … SELECT … FROM FILE, this character at the start of a line indicates a commented line to ignore.
FileReadDelimiter
,
When inserting data from a file with INSERT INTO … SELECT … FROM FILE, this character separates values If reading from a “.tsv” or “.psv” file, the default will be <tab> or |, respectively.
FileReadEscapeChar
When inserting data from a file with INSERT INTO … SELECT … FROM FILE, this character can be placed before other special characters to not treat them specially. Default is no escape character.
FileReadHasHeader
When inserting data from a file with INSERT INTO … SELECT … FROM FILE, whether to assume the first line in the file is a header line.
* 0 = The first line in the file is not a header line, treat it as data * 1 = The first line in the file is a header line, and indicates column names * If unspecified, the first line will be read and treated as a header if it looks like a header
FileReadInitialClear
0
When inserting data from a file with INSERT INTO … SELECT … FROM FILE, delete any existing data in the destination table before inserting new data.
FileReadLimit
When inserting data from a file with INSERT INTO … SELECT … FROM FILE, stop reading the file after this many lines. Default is no limit.
FileReadNullString
\N
When inserting data from a file with INSERT INTO … SELECT … FROM FILE, this string represents data that should be inserted as NULL values.
Seconds the driver will spend retrying the initial connection to the database before giving up. 0 disables retrying, in which case the URLs supplied are used as given, without auto-discovery. A non-zero value retries with a backoff that starts at one minute and doubles with each attempt. The default of -1 defers to the API default of 0. (JDBC only)
Limit
-1
Limit number of records returned for each query issued; -1 for no limit, i.e., all records will be returned for each query.
LogLevel
0
Controls the amount of debug logging done by driver.
Whether the driver will report itself as being in a read-only mode, even though it is in full-access mode; useful for 3rd party tools that expect and enforce a read-only connection.
* 0 = Driver will not report itself as being in a read-only mode when queried via the JDBC API * 1 = Driver will report itself as being in a read-only mode, even though this is not enforced
Replication
Controls the HA replication mode for all requests made through this connection.
* NONE = Do not replicate requests. * SYNC = Executes the query locally, sends an http request directly to each cluster, sequentially, and returns the response. * ASYNC = Executes the query locally, queues a request to RMQ for each cluster, and returns to the user. * SYNC_PARALLEL = Sends a http request directly to each cluster, executes the query locally, and waits for the response from each cluster. * ASYNC_PARALLEL = Queues a request to RMQ for each cluster, executes the query locally, and returns to the user.
RowsPerFetch
10000
Number of records to request from the server at a time.
RowsPerInsertion
10000
Number of records to send to the server in a batch.
Schema
Use the specified schema as the default schema for all object references
TimeZoneOverride
UTC
Time zone that client should assume time coming from the server are in. (JDBC only)
All times in Kinetica are assumed to be in UTC (GMT). Java clients should call:
TimeZone.setDefault(TimeZone.getTimeZone("UTC"));
so that times will be displayed in UTC and match other clients. When this is not possible (e.g., when using a Java utility that does not have such an option) Java clients will usually display times in the local time (set in Java and/or the OS), which will look different from the times displayed by other clients. I.e., the Java client will display the UTC times from Kinetica converted to the local time zone. The TimeZoneOverride option can be used to report times coming from Kinetica as being in the specified time zone, rather than in UTC. If this is set to the current Java/OS time zone then no conversions will be done, and displayed times should match UTC times as stored in Kinetica.
Valid options include system and time zone codes such as EST and PST.
Use system to match the current Java client time zone and avoid needing to look up the time zone code for your area.
TruncateStrings
0
Whether to truncate strings when processing an INSERT.
* 0 = Do not truncate strings — if a string is too long, it will not be inserted * 1 = Truncate strings when needed so they will fit in a restricted target string column
UseKeyLookup
0
Whether to process supported SELECT statements using high-performance distributed key lookup; using this may cause some additional overhead to determine if the table will support key lookups.
* 0 = By default, use standard processing for all SELECT statements; this can be overridden with KI_HINT_KEY_LOOKUP on individual statements * 1 = Use high-performance distributed key lookup for supported SELECT statements, when possible
SSL Parameters
The following optional parameters should be used when connecting over SSL.
Config Parameter
Default Value
Description
BypassSslCertCheck
0
Whether or not to validate the SSL certificate presented by the Kinetica server
* 0 = Do not accept SSL server certificates that fail validation checks * 1 = Accept SSL server certificates regardless of whether they pass validation checks
SslCACertPath
Path to client’s CA store, used to validate the certificate from the Kinetica server
TrustedStorePath
Path to client’s trust store, used to validate the certificate from the Kinetica server (JDBC only)
SslCACertPassword
Password to the certificate store specified by TrustedStorePath(JDBC only)
Logging
Kinetica logs inbound SQL requests, internal calls made on behalf of those
requests, and other debugging information. These logs can be found in the
following log files:
/opt/gpudb/core/logs/host-manager-<hostname>-<datetime>.log: Contains
all endpoint calls, including the initial call to /execute/sql and any
endpoint calls made as a result of servicing the request