Can I create with DynamoDB multiple tables with secondary index concurrencly?

Horcrux7_12631 0
Viewed 45

I am confused by the API documentation of CreateTable from DynamoDB. I need to create multiple tables with a secondary index. From the API: https://sdk.amazonaws.com/java/api/latest/software/amazon/awssdk/services/dynamodb/DynamoDbClient.html#createTable-software.amazon.awssdk.services.dynamodb.model.CreateTableRequest-

If you want to create multiple tables with secondary indexes on them, you must create the tables sequentially. Only one table with secondary indexes can be in the CREATING state at any given time.

and

Up to 500 simultaneous table operations are allowed per account. These operations include CreateTable, UpdateTable, DeleteTable, UpdateTimeToLive, RestoreTableFromBackup, and RestoreTableToPointInTime.

The only exception is when you are creating a table with one or more secondary indexes. You can have up to 250 such requests running at a time;

  • Can I create now only one table with a secondary index or 250 at the same time?

  • If I create multiple tables sequential without waiting on active state is this already concurrency creation?

  • Must I wait on the active state for every table if I create multiple tables with secondary indexes?

create-table amazon-dynamodb
1 Answers

An individual account can only be running one "Create Index" action at a time, no matter how many tables you have.

To understand this it may help to understand what an Index is. An Index is a complete copy of the table, but with a different partition and sort key. So if your original table has a PK of of userId and a sk of sort_key you could now create an index where the partition key is set to sort_key and the sort_key is now set to userId creating an inverted index (a common practice in Dynamo - remember Queries in Dynamo must know what the PK is, so if you have UserID you could access all data of a given User, or if you wanted all Users who have a particular tag, you may have an SK item on users that is something like TAG#ThisTag and then you wanted all users with ThisTag you could do a query against the inverted index with a pk = TAG#ThisTag and get back a list of UserIds.)

While the CreateIndex is being run on a given table, no other actions can be run on it - it wont accept changes to the data/configuration that would cause a fault/mismatch in the copying process. This is one of the reasons a given account is limited to only one create index operation at a time.

As a slight aside if I may - if you have a single account with multiple Dynamos all for the same product, you may want to rethink your database strategy. A single Dynamo Table can be used for many different storages if you set up your PK-SK as generic fields (ie: pk and sk as the attribute names) - No document inside your dynamo has to have the same attributes as any other. And when accessing data, each partition key is exactly as its named - a Partition of data that is all that is accessed when a query is made against that PK. (so if you have 100 items with PK of USER#1 and 100 items with a PK of USER#2 and you query against USER#1 you only access that 100 items - the rest are ignored by the Query and never ever touched - allowing you to in effect have multiple "tables" in a single DynamoDB Table by giving them different Partition Key prefixes.)

user_115917580
Related