Week 32, 2019 - ECS Multiple Target Groups; CloudWatch Logs Insights; PartiQL; CloudFormation Roadmap

By (4 minutes read)

Since my last weekly note, AWS has released a couple of highly anticipated features for both ECS and CloudWatch Logs Insights. In addition they open sourced a new query language and have given us a public roadmap for CloudFormation.

ECS Multiple Target Groups

Over the years ECS has grown into quite a powerful solution to manage your containers, and it’s no secret that Fargate (which falls under that) is my favourite way of running containers in the cloud. As with every solution or service however, there are things that annoy, things that make working with it a little bit less nice, and even things that mean you have to build ugly workarounds.

With ECS, one of the things in the last category was that you were limited to using only a single target group, which in turn connects to a load balancer, for a container. This meant that you couldn’t easily, for example, make the same endpoints available to an internal and an external load balancer. And if your containers aren’t exactly micro-services1 and need to listen to multiple ports, that too was impossible.

Because of that, I’m really happy to see that limitation has been lifted and you can attach multiple target groups to a task. This should definitely improve a fair number of architectures and while I have no immediate need for it I kind of wish I could go back and fix some of my old hacks2.

CloudWatch Logs Insights Cross Group

Speaking of badly needed improvements, when CloudWatch Logs Insights was introduced at re:Invent it was shown off as a way to easily query your CloudWatch logs. And in fairness, it did do that. However, it was unfortunately limited to a single log group. In case you’re not familiar with the terminology, when you configure something to use CloudWatch logs it will send logs to a specific log group. This is basically a way to organise your logs.

Being able to only search in a single log group was quite limiting however, and putting all of your logs in the same log group defeats the purpose. And with Lambda functions it’s not even possible. So, I suspect the number one feature request since the announcement was basically to support querying across multiple log groups. Which is now possible.


Speaking of querying data, AWS has released the specifications (and a reference implementation) of PartiQL3. From my understanding, PartiQL is basically an extension to SQL that allows it to deal with different types of data, many of which aren’t compatible with traditional SQL. This sounds like an interesting project, and as something that Amazon has been using internally for a while4 it will likely stay an active presence.

Of course, it is still a specification driven by a single vendor, which isn’t usually a great indicator for the likelihood of it gaining widespread adoption. That said, if you use data in the AWS ecosystem you will end up using it even if not directly.

CloudFormation Roadmap

AWS released one more thing that I’m quite happy about. Following the trend of public container roadmaps, the CloudFormation team has now also made a roadmap available on GitHub. There are quite a lot more issues in that repo than items in the roadmap dashboard, but hopefully some of those requests will be added as well.

In general though, this roadmap by definition will be less complete than others as it can’t contain unannounced services. Similarly I suspect that is also the case for unannounced features of existing services. Still, it’s good to have some idea of what’s on the roadmap regarding services and features that already exist.

Ambassador Corner

  1. Especially relevant when you use ECS to run legacy applications. [return]
  2. Somehow I doubt former employers are keen on someone coming back to “fix” something though. [return]
  3. Yes, it is indeed pronounced as ‘particle’ [return]
  4. Including not only some of their database products, but services like S3 Select as well. [return]
comments powered by Disqus