Amazon ECS-Optimized Amazon Linux AMI
Amazon Web Services | 2016.09.fLinux/Unix, Amazon Linux AMI 2016.03 - 64-bit Amazon Machine Image (AMI)
Lack of support for R4 instances.
We are currently using this product but we want to upgrade our instances to R4 instance that is memory optimized with high network performance. We are hoping that AWS will support this soon.
- Leave a Comment |
- Mark review as helpful
Simple ECS enabled amazon AMI
No frills Amazon AMI. Works as you would expect for ECS clusters. And here is some more words Amazon. Words Words Words.. What is a word?
Lacking updated Splunk driver
Almost everything you need to run your containers. I had to enable the Splunk driver on ecs.conf only to learn the driver is old, and does not support json messages (No --log-opt splunk-format=json)
Seems for a few months now, the Splunk log driver gives the ability for the log driver to pass container json logs as part of the json message to Splunk, not inline, which makes parsing easy on Splunk.
Keep that in mind if you are a Splunk user
instance-store AMI should be provided as well
Paying for EBS root device is not good if you want to run several ECS instances. If we had an instance-store version of ECS optimized we would be able to benefit from running this with EC2 Spot Fleet and using much more disk space (ephemeral0).
AMI image when executed is collabtive not Amazon Linux
Attempted to get this set up with auto scalling and ecs for the past week with the default user reported in the ui as 'root'. Had to look through the logs and found its actually collabtive linux for the this ami and the user is admin not the default AWS users: https://www.turnkeylinux.org/collabtive
This service still has a beta feel to it
Not fact that deleting(cleaning up) task definitions isn't even supported yet says it all. Strange behavior from time to time, and following the instructions on how to get the agent to pull from my private registry did not net the results I wanted. It's not all bad, however, I expected a more consistent experience for something out of beta.
The image works as expected
The image works as expected.
The root partition is too large for my usage. I think the default can be set much lower. If the deployment needs more space, they can add more when launching the instance. While shrinking the root partition is not that easy.
Definitely a good start
This is definitely a hugely desired feature for us and we are so happy that AWS did it. The potential to scale at a micro level seems to be awesome. We can go from one micro server with multiple segmented technologies in it and yet quick add little by little to the system, growing only what needs to be grown. My two biggest complaints:
1. Containers would crash/shutdown/etc with no indication of what happened (nothing in the event log or any other log that I could find). I believe it may have been a lack of resources but again I have no clue.
2. There is no registry built into AWS. I don't need the automatic build of Docker Hub and the prices are ridiculous, so I was really hoping for AWS to release an S3 backed docker registry. Charge us for the storage and a small amount for the service availability, but that shouldn't be anything near what Docker Hub charges. Plus it being directly built into AWS would allow them to create better integration in ECS.