added missing mysql root password to sample app

[abonas]

MySQL docker image requires a setting env MSQL_ROOT_PASSWORD, it doesn't work without it.
As a result pf its failure/exit, there were a lot of mysql based container instances created, and the machine ended up with "out of space", and origin was stopped as well.

[mnagy]

NACK. This is actually a different error. What happened is that I've changed the variables in #1329 but forgot to change the image from mysql to openshift/mysql-55-centos7. It should work with the image changed, can you please do that and test whether it works?

[abonas]

NACK. This is actually a different error. What happened is that I've changed the variables in #1329 but forgot to change the image from mysql to openshift/mysql-55-centos7. It should work with the image changed, can you please do that and test whether it works?

Looking at your PR #1329 seems like you're right. will wait for an official fix for that.

This whole situation raises several concerns and questions, adding @smarterclayton too.

1. Since mysql container failed to start, the system handled that by creating zillion new mysql container instances which all ended with the same error, and eventually the system ran out of space. why isn't there a limit in such situation to control the number of created containers (on k8s level perhaps?) ?
   for example, retry 5 times and that's it.
2. the implication of one missing env variable that causes the entire system to fail is quite a big issue. (origin was stopped by docker as well)
3. is there any way to prevent this in the first place before it reaches container at runtime? correlate somehow the "env" section in pod/container definition to the ones required by an image and fail the creation of a pod?
   imagine how common those errors will be, and with no validation, how hard will it be to detect and recover from those in runtime.

[mnagy]

@abonas fix here: #1762

[smarterclayton]

The fact the containers aren't cleaned up is definitely a bug - the container gc must not be working.

[mfojtik]

@smarterclayton @abonas I saw a Docker (?) bug last week on my F21 vagrant box where the space was not reclaimed and Docker filled up my partition. I tried to reboot, restart Docker, cleanup all container/images but the only solution was to rm -rf /var/lib/docker...

and i'm using native docker not lxc

[abonas]

@mfojtik mine was on fedora 20, docker version 1.5.
I removed all containers, and I also rm -rf all contents of var/lib/docker/device-mapper/metadata.
Then after I fixed the template, I have no more failed containers so it's good.
and maybe it is indeed a docker bug.

[smarterclayton]

Is this now fixed?

[mnagy]

@smarterclayton The template was fixed in #1762.

[smarterclayton]

Ok, closing then. Thanks