@mhart this is the Dockerfile.test in my repo
FROM mhart/alpine-node:6
RUN apk update && \
apk upgrade && \
apk add \
util-linux \
pciutils \
usbutils \
coreutils \
binutils \
findutils \
grep \
curl \
bash \
git
RUN apk add \
python \
py-pip \
&& \
pip install --upgrade awscli
WORKDIR /tmp/lambci/build
USER root
# Build node_modules
ADD package.json /buildserver/
RUN cd /buildserver && npm install
# Now add the rest, the above layers will be cached if package.json doesn't change
ADD . /tmp/lambci/build
# Move npm modules over
RUN rm -rf /tmp/lambci/build/server/node_modules && mv /buildserver/node_modules /tmp/lambci/build/server/
RUN cd /tmp/lambci/build && ./lambci.sh
and in lambci.sh I have a bash script that tests and builds the code. The final step if to call an aws s3 sync... which is why I need the credentials
It builds fine, the only problem is I cant deploy to s3 because I dont have access to the credentials. Not sure how to rely on the IAM role for that
When I ran this from Lambda (no Docker), it worked because the credentials are available as env variables. But I hit the 500mb limit. And in Docker the secret env vars are not passed
so you'll just need to give that role access to whatever s3 bucket you need
If you're using the CloudFormation template, you can just update the permissions from there – and update your stack: https://github.com/lambci/ecs/blob/master/cluster.template#L97-L137
@mhart I am trying to understand what controls the constraints of how many concurrent deployments are possible when using lambci/ecs.
On the CloudFormation template, the InstanceType parameter will determine the disk space/available computing power.
How about the AutoScalingGroup properties? Do the defaults mean that only 1 build can be run at the same time? If I change that do I also need to change the CreationPolicy.ResourceSignal?
On the CloudFormation template, the InstanceType parameter will determine the disk space/available computing power.
How about the AutoScalingGroup properties? Do the defaults mean that only 1 build can be run at the same time? If I change that do I also need to change the CreationPolicy.ResourceSignal?
Another thing is I have run into a "Thin pool" error, which went away once I cleaned the old Docker images. How do you deal with that, do you have a cron job on the ECS instance host that does the cleanup regularly?
I believe it is because I am building the Dockerfile.test everytime. I have seen that lambci/ecs runs
build followed by run, but it is not clear to me how to pass a shell script to the run stage. Maybe if I could leverage $LAMBCI_DOCKER_RUN_ARGS?
Having a bit of trouble just getting started with this. I attempted to use "launch stack" as well as a few different ways (all really used same template though) and I get an error on "Custom::ConfigUpdater" every time. It looks like at the point in time "Custom::ConfigUpdater" is run the ConfigTable is not created yet. I adjusted the template and added "DependsOn: ["ConfigTable"]" for "ConfigUpdater" but the result of this was a completely different error: " Failed to create resource. User: arn:aws:sts::837930595040:assumed-role/lambci-private-LambdaExecution-LR33W5LBSV98/lambci-private-build is not authorized to perform: iam:PassRole on resource: arn:aws:iam::837930595040:role/lambci-private-SnsFailures-1Q43HAMIVMMR5" . I am honestly not very familiar with CloudFormation and all the config options that seem to be available. Is nobody else having this issue with this stack?
N. Virginia
The error where the ConfigTable doesn't exist seems to not be happening now (with provided template) but it was definitely happening before =/
for eg: I want run every night not on github web hook trigger
mhart continues to already-do-things-a-year-ago that i've just now thought about.
i have a slow lambda fn that i'm thinking about moving off of lambda. but i'd like to avoid refactoring bc it's just a test and i think the reason for the move is temporary.
is anyone aware of something that'd let me ship an existing lambda function using ecs? https://github.com/lambci/ecs seems like a good start, but that's purpose-built for lambci. maybe there is something out there for just this purpose?
my goal is to continue building for lambda and as needed, move functions off lambda into ecs or back.
actually... thinking more about this. vpc support in lambda is god-awful. this would be a workaround for that too until aws gets their house in order
also efs. lol.
Is there a document that talks about where build IDs, logs, etc are stored? Also, is there any concept of visualizing the build progress or do you just watch the logs like in "classic" Jenkins?