Let's say I make a GSI for 'Name' and I have two people in my database who just happen to have the same name:
Tim Cook
Tim Cook
Now this will fail a consistency constraint on insert for duplicate values hence we need another approach.
I was thinking about hashing the name values at the end so that the BEGINS_WITH operator can still be used to search / match on but that puts you in a weird position. What do you salt with? How many characters? The longer the salt the more memory and potentially compute you waste cleaning up the salt before returning the results to the user. The shorter the salt the more likely you are to have collisions. After all there are some incredibly common names out there.
Here's an example of the values salted:
Tim Cook#ABCDEF
Tim Cook#ZYXWVU
This is great as I can insert both values now and now I can create a 'search user by name' endpoint for the user via the BEGINS_WITH('Tim Cook') operation but it feels weird.
I did a bit of searching though on sorting and searching by names in DynamoDB and didn't come up with anything meaningful on how to proceed from here. Wondering what you guys think.
My one and final issue is that names are not evenly spread out so you're inevitably going to have hotter partitions but I just don't see another way around this. Minus of course exfiltrating the data to another data store and querying it there like a full text search store.
You can’t insert to a GSI. So your concern is kind of misplaced.
You also can’t Get Item on a GSI, only Query, and that’s because there’s not necessarily one matching value for a given key.
Note: The GSI always projects the primary key over from the base table.
You can follow the following schema pattern to achieve your goal:
Partition key: Name
Sort/Range key: createdAt (The creation time of that row)
In this case, if the name is same for more than 1 people, you will be returned with all the names sorted automatically. This schema will also allow you to create a unique access pattern for each item of your table.
Partition key -> Sort key
Name -> createdAt
Tim Cook -> "HH:mm:ss"
Each row will have a different creation time and will provide unique composite key values for each item of the table.
For some reason I thought GSI's had the same uniqueness constraint as partition keys however that's not the case - you can have duplicates.
In a DynamoDB table, each key value must be unique. However, the key values in a global secondary index do not need to be unique.
Source
So a GSI is a perfectly good way to store duplicated information. Not sure this question is helpful now since it came about through ignorance so it might be worth deleting now.
Related
I am trying to create a naming convention for different objects in DynamoDB, such as tables, partition and sort keys, LSIs, GSIs, attributes, etc. I read a lot of articles and there is no common way to do that but want to learn from real-time examples to choose which one will fit best our needs.
The infrastructure I am working on is based on microservices. Along with this, some of our development environments share the same AWS account. Based on this, I ended up with something like this:
Tables: [Environment].[Service Name].[Table Name].ddb-table
GSIs/LSIs: [Environment].[Service Name].[Table Name].[GSI/LSI Name].ddb-[gsi/lsi]
Partition Key: pk ??? (in my understanding, the keys should have abstract names, because the single table stores versatile data in the same key)
Sort Key: sk ??? (in my understanding, the keys should have abstract names, because the single table stores versatile data in the same key)
Attributes: meaningful but as short as possible as they are kept for every item in the table
Different elements are separated by dot (.)
All names are separated by dashes (kebab-case) and in lower case
Tables/GSIs/LSIs are in singular form
Here is an example:
Table: dev.user-service.user-order.ddb-table
LSI: dev.user-service.user-order.lsi1pk.ddb-lsi
GSI: dev.user-service.user-order.gsi1pk.ddb-gsi
What naming conventions do you follow?
Thanks a lot in advance!
My advice:
Use PK and SK as your partition key and sort key.
Don't put table names into code. Use ParameterStore. For example, if you ever do a table restore it will be to a new table name, and if you want to send traffic to the new name you'll not want to change code.
Thus don't get too fixed to any particular table name. Never try to have code predict a table name. Only have them be consistent to help humans.
Don't put regions in your table names. When you switch to Global Tables they all keep the same name. Awkward!
GSIs can be called GSI1, GSI2, etc. GSI keys are GSI1PK and GSI1SK, etc.
Tag your tables with their name if you ever want to track per-table costs later.
Short yet meaningful attribute names are nice because it reduces storage and can reduce RCU/WCU if you're near the 4kb or 1kb lines.
Use difference accounts for dev, staging, and production. If you want to put the names into tables as well to help you spot "OMG I'm in production" that's fine.
If you have lots of attributes as the item payload which aren't used for GSIs or filtering and are always returned together, consider just storing them as a string or binary which gets parsed client side. You can even compress them. It's more efficient and lower latency because it skips the data marshaling.
For my DynamoDB table, I currently have a schema like this:
Partition key - Unique ID, so every item has a completely unique ID
Sort key - none
Attribute - JSON that contains some values
Now, I want to add a new field that will be required for every item and will indicate the specific region (e.g. NA-1, NA-2, JP-1, and so on) and I want to be able to do queries on just this field. For example, I might want to perform a query on my table to retrieve all items with the region NA-1.
My question is should I make this field a GSI? I'm new to DynamoDB so I've been researching online and it seems that using a GSI is preferred when that field may only be present for select items in the table, but my field will be required for every item, so I think using a GSI is not an option.
The other possible option I've seen is performing a scan operation and using a filter expression, but from what I've seen, that's a costly operation because DynamoDB has to look at the entire table part-by-part and then filter afterwards. My table isn't very big right now, but it may become quite large in the future, so I would like a scalable option.
TL;DR Is there someway I can add a mandatory regionID field to my table and perform efficient queries on it? What are some good options I should look into?
Yeah, a GSI might not be the best fit here. Maybe you can somehow make it part of the partition key?
Yes. Perform 2 writes on the table. First row will be what you are currently writing, and the second row will have your region as the partition key. Do not forget use transactions as it is possile that one of the writes does not succeed.
While you can use GSI, you have to realize that it is eventual consistent. It will take some time to update it and you might get inconsistent data if you query soon enough after writing.
DynamoDB is a distributed data-store i.e. it stores the data not in a single server but does partitions using the provided partition key (PK). This means your data is spread across multiple servers and brings the limitation that you can query a single partition at a time.
Coming back to your query pattern,
retrieve all items with the region X
You need to add region-id as an attribute in the main table and make it part of the GSI. Do note that to avoid conflicts you need to make the GSI SK a composite SK.
I would recommend using <region>#<unique-id>
This way you can query the GSI like,
where BEGINS_WITH ('X', SK)
Also, if any of your entry moves to a new region or a new entry is created in a region, it will automatically reflect in the GSI and your query results
I have a table structure consisting of cities and comments. I need to get all comments related to a city. I have made my primary key for comments the name of the city. Now when I query my table I can get all the comments related to the city but I need them in order of votes for that comment. the votes value are constantly changing. I have considered adding ordered by to my query or adding vote as a range key and deleting are re-adding the recored every time the votes changes. These solutions don't seam that efficient and was wondering if there is a better way of doing it?
One easy thing you could do is to use a local secondary index - this DynamoDB feature can create a second table whose hash key is the same (the city name), but the sort key is the number of votes - which remains just an ordinary attribute in your original table. DynamoDB will automatically - and consistently - take care of the second table for you as you modify the first one.
Using a LSI is easier than coding the extra deletion and addition, and more efficient in the sense of less network activity and client work - but may not be significantly cheaper in Amazon bills, because DynamoDB charges you extra for that LSI work.
I can't figure out how to do this in DynamoDB.
I have a table with data something like this:
ID Updated other fields...
1200 2017-12-11 ...
1201 2018-02-05 ...
1205 2018-01-05 ...
1206 2018-01-11 ...
1210 2018-02-15 ...
1212 2018-02-10 ...
The partition key is 'ID' and I have a sort key of 'Updated'.
I want to retrieve the records where Updated is greater than "2018-02-01", say.
I can't query on just 'Updated' alone, it complains with Query condition missed key schema element: ID. I understand what that means, but I'm not sure how to do this properly.
I've tried adding various indexes and then querying on the index, including having only the 'Updated' field as the partition key, but then I can't query for a range of values only an exact match on the partition key.
So, how do I query across multiple partitions for a condition?
I could use a scan, but that is potentially expensive. Can I do this by indexing it a certain way? Or is there a way to do something similar to a query where I don't need to specify the partition key?
Use a scan
Almost everyone using DynamoDB seems to get worried about scans. Scans are FINE in many circumstances. Things you should ask yourself include; how much data will I have, how will it grow over time, how fast do I need the scan to complete, how many RCUs will this cost? Don't just dismiss scans - do the maths.
Archive data
If you only need to access recent data, consider deleting or archiving old data. By removing it from your table you can increase the performance of scans.
Partition by date
There are various strategies you can use to improve your table performance if you really want to use a query. For example you could have a partition key of YYYY-MM and sort key of datetime (down to nanosecond). That way you can retrieve whole months of data in one query, whilst still being able to sort for specific date ranges. This kind of query is much more complicated to handle in your application than a scan. Architecting your tables really depends on your data access patterns.
Nice problem, not so nice solution! :)
• You cannot do a query without conditioning on Partition Key.
• You need the Updated column to be a Sorting Key, either in the table "schema", either in an index. If it will not be a sorting key anymore, you wont be able to efficiently query for Updated > VALUE.
So you need a constant partition key and Updated to be the sorting key. Here is your Global Secondary Index:
• PK: ConstantColumn
• SK: Updated
Of course, you'll loose some scalability because all your index will be in one partition, but using a KEYS_ONLY projection should give you enough room.
Should you really need more scalability consider having PK values like C0, C1, ..., Cn, iterate through queries for each partition key, then merge the results (divide et impera).
I would consider alternative partition keys. For example, will your business logic work if you create a GSI with year as partition key and date as sort key? How about year-month?
Your query will be more complex to write as you might have to issue multiple queries to cover more than 1 partitions to fill your result page.
But as you pointed out, this is cheaper than performing a full table scan.
I want to make sure I get this right,
Based on what I've read so far, you can NOT query a range of primary keys in dynamodb,
like if you have a primary key which is number like the phone number of your customers, you can not get items with primary keys larger than 3010000000 or between 3010000000 and 3020000000
to make it clear, I am not talking about the range key, my questions is about the primary key itself,
so if this is true, there are lots of use cases, like items between dates, users registered after some point, and... , that requiers either table scans,
is this correct?
EDIT: OK, one solution that comes to mind, would be to use only one dummy hash_key for primary key and insert the real key (like phone numbers above) as range keys, does this work?
Yes, you can not get a range of hash_key with DynamoDb. But this does not mean you are stuck with your use case.
Let's take the 'dates' use case and say your are building a logging application. You are likely to get lots of records each day.
If you use the day as the hash_key, you can put the full timestamp as the range_key. This way, you can split your query into chunks and get what you want.
Of course, to get the optimal results, you will need to know well the kind of queries. For example, what is the typical range ? With DynamoDb, as well as other key:value store, you most of the time model your data with query in mind, unlike SQL when you model with only data in mind.
Of course, if your items spans on larger/shorter range, just adapt this system.
Concerning the "all under the same dummy hash_key" sounds like a terrible idea. Sorry. I am not a hundred percent sure how it really works but I know DynamoDB does some sharding across so called partitions. I believe 1 hash_key <=> 1 partitions. Moreover, If read closely the documentation, you'll notice that the provisionned throughput is splited evenly between the partitions so that each partitions is only allocated a fraction of what you pay for.
Without modifying the keys of your primary DynamoDB table, you can add a GSI with a constant partition key and your primary table's partition key as its sort key.
This will enable you to query on the index's sort key and use the resulting partition keys to get the data you're looking for.