0

Currently we have a docker image running 24/7 on Azure Container Instance (ACI) - using the max 16GB instance. Docker image is CDC debezium-server that forwards DB changes from our on-prem DB to event-hubs. All runs fine on ACI but we are experiencing occasional OutOfMemory issues with the highest available RAM (16GB) so we are looking into something that can scale and handle these payloads better.

I was looking into:

  • Azure Container Apps (ACA) as based on the docs it should have RAM/CPU auto-scaling possibilities
  • Azure Kubernetes, but here I feel that the extra complexity is a bit downside

Please do you have any 'azure based' recommendations (also other than mentioned above) that can tackle the challenges we are having? Autoscaling to for example up to 64GB RAM with easy as possible set-up and maintenance.

miro
  • 809
  • 10
  • 25

1 Answers1

0

It seems the ACA is your best bet. I agree that AKS might be too complex for this scenario.

I recommend checking the Comparing Container Apps with other Azure container options document.

Also, I think you should test the scalability option first since the docs mentions that it is not guaranteed.

Also, keep in mind that the max CPU for a container seems to be 2 cores as mentioned here.

akathimi
  • 1,393
  • 11
  • How can ACA solve the issue since the max available configuration is 2 CPU + 4GB RAM per instance? – CSharpRocks Jan 05 '23 at 13:37
  • @CSharpRocks: [memory scaling?](https://learn.microsoft.com/en-GB/azure/container-apps/scale-app?WT.mc_id=containers-52416-stmuraws#memory) – miro Jan 05 '23 at 13:44
  • This is horizontal scaling based on memory utilization. It will add new instances, not more memory. @miro used the term memory scaling as a need to provide more memory to the instance, not scale the number of instances. – CSharpRocks Jan 05 '23 at 14:05
  • ahh I see. So ACA is a no go then :/. Just weird that in 23´we have 16GB limitations. – miro Jan 05 '23 at 19:29
  • you can run 4 replicas :). it's designed for microsevices which don't usually consume much resources. – akathimi Jan 05 '23 at 19:34