Amazon-ecs-agent: Agent Refuse to start containers with None logging driver

Created on 8 Nov 2017  路  8Comments  路  Source: aws/amazon-ecs-agent

Hi!
Followed up also here:
https://forums.aws.amazon.com/thread.jspa?threadID=267098&tstart=0
We're trying to make a deployment by running old task definition, with logging driver of None. The agents refuses to place them due to ".. is missing an attribute required by your task". When we change the log driver to syslog, it works fine.
Please fix.

Most helpful comment

Hey @ofiliz, our workaround over here was to set the Logging config. on the JSON as null at the moment we are creating a new revision.

Like this:

    ...
    "logConfiguration": null,
    ...

All 8 comments

Hi @Alonreznik. Thanks for this report. I'm looking into it.

I am affected by this since today. Every time i try to change the network mode to bridge and save, then it is automatically reset to "None".

Hi @lanoxx. In order to keep track of things, please create a new issue. This one is for tracking an apparent regression regarding logging drivers. If there's a problem with network modes, I'd very much like to keep track of that separately.

Actually, @lanoxx, your issue looks like it's probably already captured in #1057. That issue should be resolved at this time for newly launched tasks.

@Alonreznik, I'm a little confused about the actual symptoms, and I want to make sure I'm actually looking at the right things. You mention in the opening of this issue that you've got issues with the 'none' logging driver. However, forums thread that you linked to is related to network modes.

The network mode issue has been resolved within the ECS service, and I've noted as much on the forum. Are you still having logging-driver issues? If so, can you please share some more details of how you're configuring and running agent, as well as how your configuring the logging drivers in your task definition?

@lanoxx I have faced this error on AWS, after a Hard Reload (CTRL + F5 / Cmd + R) i was able to select Bridge mode and save it.

But I saw that some required attr. was added on the TD making it unable to be started on the cluster service:
com.amazonaws.ecs.capability.logging-driver.none
com.amazonaws.ecs.capability.docker-remote-api.1.18

image
^ this at the bottom of the TD detail page

Before this issue, I didn't had those attr. on my TD revisions.

@nmeyerhans assuming I faced the same issue, this is what happened to me:

  1. We had task definitions that were set 1-2+ weeks ago. These had the following requirements:
    image
  2. When clicking create new revision to update those task definitions, I reached edit container and saw this:
    image
    note: I'm not sure what was set here before and if this section existed at all. I know that we didn't set it.
  3. After creating the new revision, the requirements section changed to this:
    image
    note the added logging-driver attribute

==> Result is that I got this ... is missing an attribute required by your task error .

Workaround: I set logging-drive to syslog

Hey @ofiliz, our workaround over here was to set the Logging config. on the JSON as null at the moment we are creating a new revision.

Like this:

    ...
    "logConfiguration": null,
    ...

We've identified a problem with the ECS service, outside the agent itself, that lead to this situation. The issue has been resolved and "none" should work as a log driver again. You don't need to modify your agent in order to receive the fix.

Was this page helpful?
0 / 5 - 0 ratings

Related issues

AbelGuti picture AbelGuti  路  5Comments

dcosson picture dcosson  路  3Comments

hayajo picture hayajo  路  3Comments

leonblueconic picture leonblueconic  路  3Comments

cjbottaro picture cjbottaro  路  4Comments