5 Duties to Automate Utilizing Scheduled Question Lambdas in Rockset

Why and what to automate

As utility builders and designers, each time we see repeating duties, we instantly take into consideration the right way to automate them. This simplifies our every day work and permits us to be extra environment friendly and targeted on delivering worth to the enterprise.

Typical examples of repeating duties embody scaling compute assets to optimize their utilization from a price and efficiency perspective, sending automated e-mails or Slack messages with outcomes of a SQL question, materializing views or doing periodic copies of knowledge for growth functions, exporting knowledge to S3 buckets, and so forth.

How Rockset helps with automation

Rockset presents a set of highly effective options to assist automate widespread duties in constructing and managing knowledge options:

• a wealthy set of APIs so that each side of the platform might be managed by means of REST
• Question Lambdas – that are REST API wrappers round your parametrized SQL queries, hosted on Rockset
• scheduling of Question Lambdas – a lately launched function the place you possibly can create schedules for automated execution of your question lambdas and put up outcomes of these queries to webhooks
• compute-compute separation (together with a shared storage layer) which permits isolation and unbiased scaling of compute assets

Let’s deep dive into why these are useful for automation.

Rockset APIs permit you to work together with your entire assets – from creating integrations and collections, to creating digital cases, resizing, pausing and resuming them, to working question lambdas and plain SQL queries.

Question Lambdas supply a pleasant and straightforward to make use of method to decouple shoppers of knowledge from the underlying SQL queries so that you could hold your online business logic in a single place with full supply management, versioning and internet hosting on Rockset.

Scheduled execution of question lambdas allows you to create cron schedules that may robotically execute question lambdas and optionally put up the outcomes of these queries to webhooks. These webhooks might be hosted externally to Rockset (to additional automate your workflow, for instance to put in writing knowledge again to a supply system or ship an e-mail), however you can too name Rockset APIs and carry out duties like digital occasion resizing and even creating or resuming a digital occasion.

Compute-compute separation permits you to have devoted, remoted compute assets (digital cases) per use case. This implies you possibly can independently scale and dimension your ingestion VI and a number of secondary VIs which are used for querying knowledge. Rockset is the primary real-time analytics database to supply this function.

With the mix of those options, you possibly can automate every thing you want (besides perhaps brewing your espresso)!

Typical use circumstances for automation

Let’s now have a look into typical use circumstances for automation and present how you’ll implement them in Rockset.

Use case 1: Sending automated alerts

Typically occasions, there are necessities to ship automated alerts all through the day with outcomes of SQL queries. These might be both enterprise associated (like widespread KPIs that the enterprise is fascinated about) or extra technical (like discovering out what number of queries ran slower than 3 seconds).

Utilizing scheduled question lambdas, we will run a SQL question in opposition to Rockset and put up the outcomes of that question to an exterior endpoint resembling an e-mail supplier or Slack.

Let’s take a look at an e-commerce instance. We’ve a set known as ShopEvents with uncooked real-time occasions from a webshop. Right here we monitor each click on to each product in our webshop, after which ingest this knowledge into Rockset by way of Confluent Cloud. We’re fascinated about realizing what number of gadgets have been offered on our webshop right now and we wish to ship this knowledge by way of e-mail to our enterprise customers each six hours.

We’ll create a question lambda with the next SQL question on our ShopEvents assortment:

SELECT
    COUNT(*) As ItemsSold
FROM
    "Demo-Ecommerce".ShopEvents
WHERE 
    Timestamp >= CURRENT_DATE() AND EventType="Checkout";

We’ll then use SendGrid to ship an e-mail with the outcomes of that question. We received’t undergo the steps of establishing SendGrid, you possibly can observe that in their documentation.

When you’ve obtained an API key from SendGrid, you possibly can create a schedule in your question lambda like this, with a cron schedule of 0 */6 * * * for each 6 hours:

This may name the SendGrid REST API each 6 hours and can set off sending an e-mail with the overall variety of offered gadgets that day.

{{QUERY_ID}} and {{QUERY_RESULTS}} are template values that Rockset gives robotically for scheduled question lambdas so that you could use the ID of the question and the ensuing dataset in your webhook calls. On this case, we’re solely within the question outcomes.

After enabling this schedule, that is what you’ll get in your inbox:

You can do the identical with Slack API or some other supplier that accepts POST requests and Authorization headers and also you’ve obtained your automated alerts arrange!

In case you’re fascinated about sending alerts for gradual queries, take a look at establishing Question Logs the place you possibly can see an inventory of historic queries and their efficiency.

Use case 2: Creating materialized views or growth datasets

Rockset helps automated real-time rollups on ingestion for some knowledge sources. Nonetheless, when you’ve got a must create further materialized views with extra complicated logic or if it’s essential have a duplicate of your knowledge for different functions (like archival, growth of recent options, and so forth.), you are able to do it periodically by utilizing an INSERT INTO scheduled question lambda. INSERT INTO is a pleasant method to insert the outcomes of a SQL question into an present assortment (it might be the identical assortment or a totally totally different one).

Let’s once more take a look at our e-commerce instance. We’ve an information retention coverage set on our ShopEvents assortment in order that occasions which are older than 12 months robotically get faraway from Rockset.

Nonetheless, for gross sales analytics functions, we wish to make a copy of particular occasions, the place the occasion was a product order. For this, we’ll create a brand new assortment known as OrdersAnalytics with none knowledge retention coverage. We’ll then periodically insert knowledge into this assortment from the uncooked occasions assortment earlier than the info will get purged.

We are able to do that by making a SQL question that can get all Checkout occasions for the day before today:

INSERT INTO "Demo-Ecommerce".OrdersAnalytics
SELECT
    e.EventId AS _id,
    e.Timestamp, 
    e.EventType, 
    e.EventDetails, 
    e.GeoLocation, 
FROM
    "Demo-Ecommerce".ShopEvents e
WHERE 
    e.Timestamp BETWEEN CURRENT_DATE() - DAYS(1) AND CURRENT_DATE()
    AND e.EventType="Checkout";

Be aware the _id area we’re utilizing on this question – this can make sure that we don’t get any duplicates in our orders assortment. Take a look at how Rockset robotically handles upserts right here.

Then we create a question lambda with this SQL question syntax, and create a schedule to run this as soon as a day at 1 AM, with a cron schedule 0 1 * * *. We don’t must do something with a webhook, so this a part of the schedule definition is empty.

That’s it – now we’ll have every day product orders saved in our OrdersAnalytics assortment, prepared to be used.

Use case 3: Periodic exporting of knowledge to S3

You should utilize scheduled question lambdas to periodically execute a SQL question and export the outcomes of that question to a vacation spot of your selection, resembling an S3 bucket. That is helpful for situations the place it’s essential export knowledge regularly, resembling backing up knowledge, creating studies or feeding knowledge into downstream programs.

On this instance, we’ll once more work on our e-commerce dataset and we’ll leverage AWS API Gateway to create a webhook that our question lambda can name to export the outcomes of a question into an S3 bucket.

Much like our earlier instance, we’ll write a SQL question to get all occasions from the day before today, be part of that with product metadata and we’ll save this question as a question lambda. That is the dataset we wish to periodically export to S3.

SELECT
    e.Timestamp, 
    e.EventType, 
    e.EventDetails, 
    e.GeoLocation, 
    p.ProductName, 
    p.ProductCategory, 
    p.ProductDescription, 
    p.Value
FROM
    "Demo-Ecommerce".ShopEvents e
    INNER JOIN "Demo-Ecommerce".Merchandise p ON e.EventDetails.ProductID = p._id
WHERE 
    e.Timestamp BETWEEN CURRENT_DATE() - DAYS(1) AND CURRENT_DATE();

Subsequent, we’ll must create an S3 bucket and arrange AWS API Gateway with an IAM Position and Coverage in order that the API gateway can write knowledge to S3. On this weblog, we’ll give attention to the API gateway half – remember to test the AWS documentation on the right way to create an S3 bucket and the IAM function and coverage.

Comply with these steps to arrange AWS API Gateway so it’s prepared to speak with our scheduled question lambda:

1. Create a REST API utility within the AWS API Gateway service, we will name it rockset_export:

2. Create a brand new useful resource which our question lambdas will use, we’ll name it webhook:

3. Create a brand new POST methodology utilizing the settings under – this basically allows our endpoint to speak with an S3 bucket known as rockset_export:

• AWS Area: Area in your S3 bucket
• AWS Service: Easy Storage Service (S3)
• HTTP methodology: PUT
• Motion Kind: Use path override
• Path override (elective): rockset_export/{question _id} (substitute together with your bucket identify)
• Execution function: arn:awsiam::###:function/rockset_export (substitute together with your ARN function)
• Setup URL Path Parameters and Mapping Templates for the Integration Request – this can extract a parameter known as query_id from the physique of the incoming request (we’ll use this as a reputation for recordsdata saved to S3) and query_results which we’ll use for the contents of the file (that is the results of our question lambda):

As soon as that’s achieved, we will deploy our API Gateway to a Stage and we’re now able to name this endpoint from our scheduled question lambda.

Let’s now configure the schedule for our question lambda. We are able to use a cron schedule 0 2 * * * in order that our question lambda runs at 2 AM within the morning and produces the dataset we have to export. We’ll name the webhook we created within the earlier steps, and we’ll provide query_id and query_results as parameters within the physique of the POST request:

We’re utilizing {{QUERY_ID}} and {{QUERY_RESULTS}} within the payload configuration and passing them to the API Gateway which is able to use them when exporting to S3 because the identify of the file (the ID of the question) and its contents (the results of the question), as described in step 4 above.

As soon as we save this schedule, we’ve an automatic process that runs each morning at 2 AM, grabs a snapshot of our knowledge and sends it to an API Gateway webhook which exports this to an S3 bucket.

Use case 4: Scheduled resizing of digital cases

Rockset has assist for auto-scaling digital cases, but when your workload has predictable or nicely understood utilization patterns, you possibly can profit from scaling your compute assets up or down primarily based on a set schedule.

That means, you possibly can optimize each spend (so that you simply don’t over-provision assets) and efficiency (so that you’re prepared with extra compute energy when your customers wish to use the system).

An instance might be a B2B use case the place your prospects work primarily in enterprise hours, let’s say 9 AM to five PM all through the work days, and so that you want extra compute assets throughout these occasions.

To deal with this use case, you possibly can create a scheduled question lambda that can name Rockset’s digital occasion endpoint and scale it up and down primarily based on a cron schedule.

Comply with these steps:

1. Create a question lambda with only a choose 1 question, since we don’t really need any particular knowledge for this to work.
2. Create a schedule for this question lambda. In our case, we wish to execute as soon as a day at 9 AM so our cron schedule will probably be 0 9 * * * and we’ll set limitless variety of executions in order that it runs day-after-day indefinitely.
3. We’ll name the replace digital occasion webhook for the precise VI that we wish to scale up. We have to provide the digital occasion ID within the webhook URL, the authentication header with the API key (it wants permissions to edit the VI) and the parameter with the NEW_SIZE set to one thing like MEDIUM or LARGE within the physique of the request.

We are able to repeat steps 1-3 to create a brand new schedule for scaling the VI down, altering the cron schedule to one thing like 5 PM and utilizing a smaller dimension for the NEW_SIZE parameter.

Use case 5: Organising knowledge analyst environments

With Rockset’s compute-compute separation, it’s straightforward to spin up devoted, remoted and scalable environments in your advert hoc knowledge evaluation. Every use case can have its personal digital occasion, making certain {that a} manufacturing workload stays secure and performant, with the most effective price-performance for that workload.

On this situation, let’s assume we’ve knowledge analysts or knowledge scientists who wish to run advert hoc SQL queries to discover knowledge and work on numerous knowledge fashions as a part of a brand new function the enterprise desires to roll out. They want entry to collections they usually want compute assets however we don’t need them to create or scale these assets on their very own.

To cater to this requirement, we will create a brand new digital occasion devoted to knowledge analysts, make sure that they will’t edit or create VIs by making a customized RBAC function and assign analysts to that function, and we will then create a scheduled question lambda that can resume the digital occasion each morning in order that knowledge analysts have an atmosphere prepared once they log into the Rockset console. We might even couple this with use case 2 and create a every day snapshot of manufacturing right into a separate assortment and have the analysts work on that dataset from their digital occasion.

The steps for this use case are much like the one the place we scale the VIs up and down:

1. Create a question lambda with only a choose 1 question, since we don’t really need any particular knowledge for this to work.
2. Create a schedule for this question lambda, let’s say every day at 8 AM Monday to Friday and we’ll restrict it to 10 executions as a result of we would like this to solely work within the subsequent 2 working weeks. Our cron schedule will probably be 0 8 * * 1-5.
3. We’ll name the resume VI endpoint. We have to provide the digital occasion ID within the webhook URL, the authentication header with the API key (it wants permissions to renew the VI). We don’t want any parameters within the physique of the request.

That’s it! Now we’ve a working atmosphere for our knowledge analysts and knowledge scientists that’s up and working for them each work day at 8 AM. We are able to edit the VI to both auto-suspend after sure variety of hours or we will have one other scheduled execution which is able to droop the VIs at a set schedule.

As demonstrated above, Rockset presents a set of helpful options to automate widespread duties in constructing and sustaining knowledge options. The wealthy set of APIs mixed with the facility of question lambdas and scheduling permit you to implement and automate workflows which are utterly hosted and working in Rockset so that you simply don’t should depend on third get together parts or arrange infrastructure to automate repeating duties.

We hope this weblog gave you a couple of concepts on the right way to do automation in Rockset. Give this a attempt to tell us the way it works!