saveToCassandra specify column names in Spark Cassandra Connector - cassandra

I'm using Spark Cassandra Connector to update counters in this table:
CREATE TABLE IF NOT EXISTS analytics.minute_usage_stats (
metric_date timestamp,
user_id uuid,
metric_name text,
h1_m1 counter,
h1_m2 counter,
h1_m3 counter,
...
h24_m60 counter,
PRIMARY KEY ((metric_date), user_id, metric_name)
);
The code looks something like this:
class Metric(metricName: String, metricDate: Date, userId: String, ???)
metricEventStream
.map(event => {
// Parsing logic
new Metric(metricName, metricDate, userId, h1M1???)
})
.saveToCassandra("analytics", "minute_usage_stats")
I need to be able to update one counter column at a time and the columns could be different for each item in the map.
Looking through the documentation, there doesn't seem to be a way to do this without using the Cassandra driver itself and losing all the great functionality of the Spark connector.
Is there any way to specify column names per RDD row with the current Spark Cassandra Connector?

Related

Hive, how to partition by a colum with null values, putting all nulls in one partition

I am using Hive, and the IDE is Hue. I am trying different key combinations to choose for my partition key(s).
The definition of my original table is as follows:
CREATE External Table `my_hive_db`.`my_table`(
`col_id` bigint,
`result_section__col2` string,
`result_section_col3` string ,
`result_section_col4` string,
`result_section_col5` string,
`result_section_col6__label` string,
`result_section_col7__label_id` bigint ,
`result_section_text` string ,
`result_section_unit` string,
`result_section_col` string ,
`result_section_title` string,
`result_section_title_id` bigint,
`col13` string,
`timestamp` bigint,
`date_day` string
)
PARTITIONED BY (
`date_year` string,
`date_month` string)
ROW FORMAT SERDE
'org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe'
STORED AS INPUTFORMAT
'org.apache.hadoop.hive.ql.io.parquet.MapredParquetInputFormat'
OUTPUTFORMAT
'org.apache.hadoop.hive.ql.io.parquet.MapredParquetOutputFormat'
LOCATION
's3a://some/where/in/amazon/s3';
The above code is working properly. But when I create a new table with date_day as partition key, the table is empty and I need to run MSCK Repair Table. However, I am getting the following error:
Error while compiling statement: FAILED: Execution Error, return code 1 from org.apache.hadoop.hive.ql.ddl.DDLTask
When the partion keys were date_year, date_month, MSCK worked properly.
Table definition of the table I am getting the error for is as follows:
CREATE External Table `my_hive_db`.`my_table`(
`col_id` bigint,
`result_section__col2` string,
`result_section_col3` string ,
`result_section_col4` string,
`result_section_col5` string,
`result_section_col6__label` string,
`result_section_col7__label_id` bigint ,
`result_section_text` string ,
`result_section_unit` string,
`result_section_col` string ,
`result_section_title` string,
`result_section_title_id` bigint,
`col13` string,
`timestamp` bigint,
`date_year` string,
`date_month` string
)
PARTITIONED BY (
`date_day` string)
ROW FORMAT SERDE
'org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe'
STORED AS INPUTFORMAT
'org.apache.hadoop.hive.ql.io.parquet.MapredParquetInputFormat'
OUTPUTFORMAT
'org.apache.hadoop.hive.ql.io.parquet.MapredParquetOutputFormat'
LOCATION
's3a://some/where/in/amazon/s3';
After this the following query is empty:
Select * From `my_hive_db`.`my_table` Limit 10;
I therefore ran the following command:
MSCK REPAIR TABLE `my_hive_db`.`my_table`;
And I get the error: Error while compiling statement: FAILED: Execution Error, return code 1 from org.apache.hadoop.hive.ql.ddl.DDLTask
I checked this link as it is exactly the error I am getting, but by using the solution provided:
set hive.msck.path.validation=ignore;
MSCK REPAIR TABLE table_name;
I get a different error:
Error while processing statement: Cannot modify hive.msck.path.validation at runtime. It is not in list of params that are allowed to be modified at runtime.
I think the reason I am getting these errors is that there are more than 200 million records with date_day having null value.
There are 31 distinct date-day not null values. I would like to partition my table with 32 partitions, each for a distinct value of date_day field, and all the null values get into a different partition. Is there a way to do so (partitioning by a column with null values)?
If this can be achieved by spark, I am also open to use it.
This is part of a bigger problem of changing partition keys by recreating a table as mentioned in this link in answer to my other question.
Thank you for your help.
You seem to not understand how Hive's partitioning work.
Hive stores data into files on HDFS (or S3, or some other distributed folders).
If you create a non-partitioned parquet table called my_schema.my_table, you will see in your distributed storage files stored in a folder
hive/warehouse/my_schema.db/my_table/part_00001.parquet
hive/warehouse/my_schema.db/my_table/part_00002.parquet
...
If you create a table partitioned by a column p_col, the files will look like
hive/warehouse/my_schema.db/my_table/p_col=value1/part_00001.parquet
hive/warehouse/my_schema.db/my_table/p_col=value1/part_00002.parquet
...
hive/warehouse/my_schema.db/my_table/p_col=value2/part_00001.parquet
hive/warehouse/my_schema.db/my_table/p_col=value2/part_00002.parquet
...
The command MSCK repair table allows you to automatically reload the partitions, when you create an external table.
Let's say you have folders on s3 that look like this:
hive/warehouse/my_schema.db/my_table/p_col=value1/part_00001.parquet
hive/warehouse/my_schema.db/my_table/p_col=value2/part_00001.parquet
hive/warehouse/my_schema.db/my_table/p_col=value3/part_00001.parquet
You create an external table with
CREATE External Table my_schema.my_table(
... some columns ...
)
PARTITIONED BY (p_col STRING)
the table will be created but empty, because Hive hasn't detected the partitions yet. You run MSCK REPAIR TABLE my_schema.my_table, and Hive will recognize that your partition p_col matches the partitioning scheme on s3 (/p_col=value1/).
From what I could understand from your other question, you are trying to change the partitioning scheme of the table by doing
CREATE External Table my_schema.my_table(
... some columns ...
)
PARTITIONED BY (p_another_col STRING)
and you are getting an error message because p_another_col doesn't match with the column used in s3, which was p_col.
And this error is perfectly normal, since what you are doing doesn't make sense.
As stated in the other question's answer, you need to create a copy of the first table, with a different partitioning scheme.
You should instead try something like this:
CREATE External Table my_hive_db.my_table_2(
`col_id` bigint,
`result_section__col2` string,
`result_section_col3` string ,
`result_section_col4` string,
`result_section_col5` string,
`result_section_col6__label` string,
`result_section_col7__label_id` bigint ,
`result_section_text` string ,
`result_section_unit` string,
`result_section_col` string ,
`result_section_title` string,
`result_section_title_id` bigint,
`col13` string,
`timestamp` bigint,
`date_year` string,
`date_month` string
)
PARTITIONED BY (`date_day` string)
and then populate your new table with dynamic partitioning
INSERT OVERWRITE TABLE my_hive_db.my_table_2 PARTITION(date_day)
SELECT
col_id,
result_section__col2,
result_section_col3,
result_section_col4,
result_section_col5,
result_section_col6__label,
result_section_col7__label_id,
result_section_text,
result_section_unit,
result_section_col,
result_section_title,
result_section_title_id,
col13,
timestamp,
date_year,
date_month,
date_day
FROM my_hive_db.my_table_1

Reading guarantees for full table scan while updating the table?

Given schema:
CREATE TABLE keyspace.table (
key text,
ckey text,
value text
PRIMARY KEY (key, ckey)
)
...and Spark pseudocode:
val sc: SparkContext = ...
val connector: CassandraConnector = ...
sc.cassandraTable("keyspace", "table")
.mapPartitions { partition =>
connector.withSessionDo { session =>
partition.foreach { row =>
val key = row.getString("key")
val ckey = Random.nextString(42)
val value = row.getString("value")
session.execute(s"INSERT INTO keyspace.table (key, ckey, value)" +
" VALUES ($key, $ckey, $value)")
}
}
}
Is it possible for a code like this to read an inserted value within a single application (Spark job) run? More generalized version of my question would be whether a token range scan CQL query can read newly inserted values while iterating over rows.
Yes, it is possible exactly as Alex wrote
but I don't think it's possible with above code
So per data model the table is ordered by ckey in ascending order
The funny part however is the page size and how many pages are prefetched and since this is by default 1000 (spark.cassandra.input.fetch.sizeInRows), then the only problem could occur, if you wouldn't use 42, but something bigger and/or the executor didn't page yet
Also I think you use unnecessary nesting, so the code to achieve what you want might be simplified (after all cassandraTable will give you a data frame).
(I hope I understand that you want to read per partition (note a partition in your case is all rows under one primary key - "key") and for every row (distinguished by ckey) in this partition generate new one (with new ckey that will just duplicate value with new ckey) - use case for such code is a mystery for me, but I hope it has some sense:-))

CQL table design for temporal data

As a Cassandra novice, I have a CQL design question. I want to re-use a concept which I've build before using RDBMS systems, to create history for customerData. The customer himself will only see the latest version, so that should be the fastest, but queries on whole history can be performed.
My suggested entity properties:
customerId text,
validFromDate date,
validUntilDate date,
customerData text
First save of customerData just INSERTs customerData with validFromDate=NOW and validUntilDate=31-12-9999
Subsequent saves of customerData changes the last record - setting validUntilDate=NOW - and INSERT new customerData with validFromDate=NOW and validUntilDate=31-12-9999
Result:
This way a query of (customerId, validUntilDate)=(id,31-12-9999) will give last saved version.
Query on (customerId) will give all history.
To query customerData at certain time t just use query with validFromDate < t < validUntilDate
My guess is PARTITION_KEY = customerId and CLUSTER_KEY can be validFromDate. Or use PRIMARY KEY = customerId. Or I could create two tables, one for fast querying of lastest version (has no history), and another for historical analyses.
How do you design this in CQL-way? I think I'm thinking too much RDBMish.
Use change timestamp as CLUSTERING KEY with DESC order, e.g
CREATE TABLE customer_data_versions (
id text,
change_time timestamp,
name text,
PRIMARY KEY (id, change_time)
) WITH CLUSTERING ORDER BY ( change_time DESC );
It will allow you to store data versions per customer id in descending order.
Insert two versions for the same id:
INSERT INTO customer_data_versions (id, change_time, name) VALUES ('id1', totimestamp(now()),'John');
INSERT INTO customer_data_versions (id, change_time, name) VALUES ('id1', totimestamp(now()),'John Doe');
Get last saved version:
SELECT * FROM customer_data_versions WHERE id='id1' LIMIT 1;
Get all versions for the id:
SELECT * FROM customer_data_versions WHERE id='id1';
Get versions between dates:
SELECT * FROM customer_data_versions WHERE id='id1' AND change_time <= before_date AND change_time >= after_date;
Please note, there are some limits for partition size (how much versions you will be able to store per customer id):
Cells in a partition: ~2 billion (231); single column value size: 2 GB ( 1 MB is recommended)

How to select Cluster keys that are in the same partition key?

I'm creating a like oneToMany relationShip with cassandra CQL 3,
A service may have different responses, and response belong to one service.
I would like to know if there's a function that can let me select Cluster keys that are in the same partition key?
CREATE TABLE reponse (
service_id uuid,
service_nom text static,
service_timeout boolean static,
reponse_id uuid,
reponse_type boolean,
reponse_nom text,
reponse_valide boolean,
reponse_message text,
PRIMARY KEY((service_id), reponse_type, reponse_id )
);
Thanks in advance!

Spark Cassandra Connector keyBy and shuffling

I am trying to optimize my spark job by avoiding shuffling as much as possible.
I am using cassandraTable to create the RDD.
The column family's column names are dynamic, thus it is defined as follows:
CREATE TABLE "Profile" (
key text,
column1 text,
value blob,
PRIMARY KEY (key, column1)
) WITH COMPACT STORAGE AND
bloom_filter_fp_chance=0.010000 AND
caching='ALL' AND
...
This definition results in CassandraRow RDD elements in the following format:
CassandraRow <key, column1, value>
key - the RowKey
column1 - the value of column1 is the name of the dynamic column
value - the value of the dynamic column
So if I have RK='profile1', with columns name='George' and age='34', the resulting RDD will be:
CassandraRow<key=profile1, column1=name, value=George>
CassandraRow<key=profile1, column1=age, value=34>
Then I need to group elements that share the same key together to get a PairRdd:
PairRdd<String, Iterable<CassandraRow>>
Important to say, that all the elements I need to group are in the same Cassandra node (share the same row key), so I expect the connector to keep the locality of the data.
The problem is that using groupBy or groupByKey causes shuffling. I rather group them locally, because all the data is on the same node:
JavaPairRDD<String, Iterable<CassandraRow>> rdd = javaFunctions(context)
.cassandraTable(ks, "Profile")
.groupBy(new Function<ColumnFamilyModel, String>() {
#Override
public String call(ColumnFamilyModel arg0) throws Exception {
return arg0.getKey();
}
})
My questions are:
Does using keyBy on the RDD will cause shuffling, or will it keep the data locally?
Is there a way to group the elements by key without shuffling? I read about mapPartitions, but didn't quite understand the usage of it.
Thanks,
Shai
I think you are looking for spanByKey, a cassandra-connector specific operation that takes advantage of the ordering provided by cassandra to allow grouping of elements without incurring in a shuffle stage.
In your case, it should look like:
sc.cassandraTable("keyspace", "Profile")
.keyBy(row => (row.getString("key")))
.spanByKey
Read more in the docs:
https://github.com/datastax/spark-cassandra-connector/blob/master/doc/3_selection.md#grouping-rows-by-partition-key

Resources