Test questions
Youtube
1
Only single ((eg in File_Format) and double quotes (identifiers) are allowed as value enclosures
Comments are -- or // (probably for Javascript)
12 - Use STRIP_OUTER_ELEMENT to remove XML root element
20 - Only the Data Cache (Warehouse Cache) can be controlled by the user (by suspending WH thus clearing it)
21 - Data load rate is affected by: physical location (cross-region etc) and compression efficiency (high IO if uncompressed)
48 - Variant Types: 1) Optimized storage based on repeated elements 2) can be queried using JSON path notation
50 - Streams can be created on temporary and external tables
51 - A task can have only max of 100 child tasks
52 - Query Statements encryption is supported on BC edition
53 - MFA can be used with JDBC driver
2
AccountAdmin and SecurityAdmin can create network policies
Tech Ecosystem partner categories: ALL (Data Integration, BI, SQL Editors, Security & Governance, Advanced Analytics, Programmatic Interface)
Type of stream allowed on external table: Insert Only
Tasks tree can only refer to other tasks in the same schema
When a Task is created initially, it's state is Suspended
BC edition encrypts all data, including within the VPC
Triple redundancy for critical Snowflake services
Resource Monitors: can suspend warehouse at specific date and time regardless of credit used
3
Binary data types can be of max 8MB
While cloning schema or database, pipes that refer external stage are cloned and the ones that reference internal stage are not cloned
VARCHAR without length is same as VARCHAR with max length
Only Enterprise+ accounts can opt for early release access
Resource Monitors with "Suspend Immediate" may still incur charges while the VWH is being suspended
Snowflake's full release consists of 3 stages spread out over 2 days, i.e. stage 1, all early access stage 2: all standard stage 3: all enterprise+.
state 1 happens on day 1, Stage 2 may happen on day 1 or 2, stage 3 happens on day 2
patch release moved to all customers on the same day
converting char to boolean: (false, 0, no) all abbreviations and cases
Only AccountAdmin can create a SHARE
Standard views cannot be shared, only secured views can be
4
set auto_suspend=NULL (not 0) to ensure WH never suspends
Internal named stage cannot be cloned
Use WAREHOUSE_METERING_HISTORY in either information_schema or account_usage, or Web UI to find out credit used
Sharing data from/to VPS edition isn't supported
SQLs start running after all servers have been provisioned within a warehouse, but on a server failure, Snowflake starts running SQL when at least 50% of the servers are available
Load metadata expire after 64 days
Snowpipe history is stored in Pipe's (as opposed to table's) metadata and available for 14 days
Experiment with WH sizes (as opposed to scaling up or down until you don't see any changes)
altering a WH down affects only when there are no SQLs are running on the warehouse
5
COPY INTO command has validation_mode = 'return_errors' that allows checking data without actually loading in
WH can be resized even if it's suspended
Federated authentication and MFA are available to all editions
Uncompressed size of a micro-partition is 50-500MB
COPY INTO can transform semi-structured data during load
EXECUTE TASK account level privilege is needed by the task owning role in addition to perform the actual operation being performed in the task
When cloning schema or database, the contained objects retain privileges of the source objects
Udemy
1
6) Cloud services provide Infrastructure Management in addition to Metadata Management and Authentication
31) Can't grant ownership of an existing share to some other role
32) NULLs are sorted higher than the other values, (sorted by numeric byte value of each character for ASCII tables?)
52) Clones can load files again from stages => TRUE
62) An idle query holding locks will abort in 4 hours if blocking, 5 minutes otherwise
72) INFORMATION_SCHEMA table functions can be used to obtain account-level information such as storage, warehouses and queries
73) Queries against INFORMATIOIN_SCHEMA views will return error if they return too much data
75) MAX_FILE_SIZE to limit COPY INTO file size
79) COPY INTO allows unloading to S3 locations besides internal and external stages
82) GET supports PARALLEL option to use multiple threads
94) Snowpipe can load file again if modified => FALSE because de-dup works only on filename
Bulk-copy de-dup works on filename+MD5
95) Snowpipe REST API works with both, internal and external stages
2
12) Commands to find out if a stream has become stale: SHOW STREAM or DESCRIBE STREAM
13) TASK_HISTORY view required permission: ownership or monitor or operate on task, monitor execution, accountadmin
16) Only the table owner can create initial stream, additional streams require USAGE on database and schema, CREATE STREAM and SELECT on table
39) COPY INTO with PURGE=TRUE doesn't report error if the file has been successfully loaded but cannot be removed for some reason
44) COPY INTO <location> requires s3:PutObject and s3:DeleteObject permissions, but s3:ListObject is not required
59) system$stream_get_table_timestamp gets current offset
62) For a multi-threaded application use , synchronous queries with separate connection per thread instead of asynchronous queries with single connection shared between threads
73) SI parameter name that limits locations that can be accessed: STORAGE_ALLOWED_LOCATIONS
89) To add clustering, USAGE or ownership privileges on schema and database are required in addition to ownership privilege on the table
91) Default USER_TASK_TIMEOUT_MS is 1 hour before Snowflake kills a long running task
93) Cloning a schema that contains table and a stream, will cause the unconsumed records in cloned stream to be inaccessible
99) create storage integration required params: name, type, enabled, storage_allowed_locations
3
5) Partner categories: Data Integration, Business Intelligence, Security & Governance and ML and data science
24) Time travel cannot be disabled at account level, but can be at database, schema and table levels
25) RESOURCE MONITOR can be reassigned from account-level to warehouse-level and vice-versa
36) Cannot use time-travel with shares
57) Continuous data loading options: Snowpipe, Kafka-connector and third-party DI tools
58) Neither Streams nor time-travel are supported for materialized views
64) PIPE_EXECUTION_PAUSED can be set at Account, Schema and Pipe level (but not Database)
73) Empty table has a clustering depth of zero
74) range predicates on high-cardinality columns that are clustered will perform poorly
75) Temporary tables can have clustering key
77) Parquet file size limit is 1GB
4
12) HIPAA requires signed agreement between customer and Snowflake before storing PHI data
5
70) Data compression applies only to CSV and JSON: FALSE, Parquet can also specify compression
Udemy AA
1
PIPE doesn't support: FILES=(..), MATCH_BY_COLUMN_NAME, FORCE, PURGE, ON_ERROR=ABORT_STATEMENT, VALIDATION_MODE, RETURN_FAILED_ONLY
MCW, new cluster started by default: immediately either query is queued or there's one more query than current clusters can execute
Who can grant permission EXECUTE TASK? It's a global privilege, so only ACCOUNTADMIN
NULLs are ordered higher than any other values; use NULL FIRST/LAST to override
loadHistoryScan should not be called too quickly and should be narrow; e.g. read the last 10 minutes every 8 minutes
name of the parameter to disable query cache: USE_CACHED_RESULT
COPY INTO for unloading supports all SQLs including JOIN
Extending time-travel: 1) any data in fail-safe has no impact 2) any data already in time-travel will be extended
Snowpipe doesn't de-dup by eTag or MD5, just the filename
2
Snowpipe insertFiles supports 1) up to 5000 files, 2) file path <= 1024 when serialized to UTF-8
Make files on external stage available with low latency: use Snowpipe (not MV)
VALIDATION_MODE: RETURN_ALL_ERRORS
COMPRESSION=AUTO supports all except BROTLI
Can't create clustering keys on VARIANT but you can create on a variant path, and also on expressions
Unload permissions needed: s3:DeleteObject and s3:PutObject
Maximum task run-time: 60 minutes before it's cancelled
streams cannot be created on materialized views
3
Cloning a PIPE will cause Snowpipe to load files again into the source table
revoke USAGE, SELECT access with CASCADE to prevent your owned objects from being shared by the share owner
Snowflake collects stats on parts of VARIANT data, so VARIANT columns can be used for pruning
4
By default the max unload file size is 16MB
Privileges on objects are not replicated
5
PUT on GCP always overwrites
6
Snowflake replicates data to three availability zones
Compression applies to, besides JSON and CSV, parquet files too
Udemy PT
1
WH credit usage: INFORMATION_SCHEMA.WAREHOUSE_METERING_HISTORY and ACCOUNT_USAGE.WAREHOUSE_METERING_HISTORY
Use Account Tab for warehouse usage (not Warehouse tab)
2
COPY transformations include concatenation of columns
Both TO_NUMBER and TO_NUMERIC are valid casting functions
Cloning