Howdy, is it intentional that the docker images fo...
# all-things-deployment
s
Howdy, is it intentional that the docker images for the consumer jobs are using alpine linux? https://github.com/linkedin/datahub/blob/master/docker/datahub-mce-consumer/Dockerfile#L4 Specifically running into errors like this because alpine linux does not use glibc. I think this error is only showing up in our real kafka environments/not docker compose because our kafka cluster is configured to use snappy compression by default.
Copy code
java.lang.UnsatisfiedLinkError: /tmp/snappy-1.1.7-bb847a5e-21b5-4d9b-babd-f31afc7109a7-libsnappyjava.so: Error loading shared library ld-linux-x86-64.so.2: No such file or directory (needed by /tmp/snappy-1.1.7-bb847a5e-21b5-4d9b-babd-f31afc7109a7-libsnappyjava.so)
	at java.lang.ClassLoader$NativeLibrary.load(Native Method)
s
That was an intentional change from
openjdk:8
to
openjdk:8-jre-alpine
to reduce the Docker image size
I think you can always build your custom images by forking it as a solution but before going into that route, it's not clear to me why you need
glibc
for consumers?
s
We have snappy compression enabled on all of our kafka topics by default. It looks like the kafka consumer is using a snappy native library to do the decompression from the topic, and that native library is linked against glibc
There might be a way to install a compat layer for libc into alpine as described here: https://github.com/gliderlabs/docker-alpine/issues/219 I'll let you know if it works.
I think that libc6-compat layer worked, still working through some other issues once I got past the snappy library loading issues. Any issues if I open a PR adding that for the consumers (after some more testing)?
s
Thanks for working that out. We always appreciate contribution 🙂 Waiting for your PR.
s
Did not forget about this, just trying to get schema registry/ssl sorted out so I can do an end to end test. Will prob open a PR early next week.
👍 1
Finally opened a PR for this, not sure if this is the best approach I gave a possible alternative in the PR: https://github.com/linkedin/datahub/pull/1899/files