-
-
Notifications
You must be signed in to change notification settings - Fork 867
Description
Concourse 3.5.0 Docker images don't work on CentOS 7, kernel 4.12.9-1.el7.elrepo.x86_64
, Docker 17.05.0-ce, build 89658be
, run within Docker Swarm.
runc create: exit status 1: container_linux.go:264: starting container process caused "process_linux.go:247: running exec setns process for init caused \"exit status 38\""
Considering #1424, #1501, and the fact that I've had serious problems related to containers and volumes with almost every version of Concourse on a variety of environments for the past few months of versions (usually leaky containers, getting stuck after a few hours or days, or failing to start entirely), without wanting to sound unduly confrontational (or ungrateful)—on which environments does Concourse work through the Docker images? Pre 3.5.0, it at least ran on the described environment with a later-than-standard kernel, albeit periodically breaking every few days and needing the workers to be recreated, and experiencing frequent stalled workers and volumes reaped from underneath—even if it didn't work on any recent Kubernetes cluster I've been running, no matter what the kernel. But now, even the configuration which did work is broken, and now I need to roll back to 3.4.1 again (which also has plenty of problems). Is there an environment build matrix or something somewhere? As I say, I don't want to sound unduly confrontational, here—but having experienced so many problems (such as the ones I've reported previously) on a variety of installations, OSes, kernels, and image orchestrators, this is becoming frustrating. Is Concourse 3.x stable? :|
Update: I'm suspecting this is another kernel-related incompatibility; the kernel was recently upgraded within the channel I'm tracking to a later 4.12
, and having rolled back to Concourse 3.4.1, this doesn't appear to work anymore either. So, it might not actually be Concourse 3.5.0 itself, but rather a wider issue.
Update: This also appears to be a problem with kernel 4.13.3-1.el7.elrepo.x86_64
.
Update: This also appears to be a problem with stock kernel 3.10.0-693.2.2.el7.x86_64
. In case you're wondering why I'm usually tracking mainline 4.x
above, it's because on previous versions of Concourse workers didn't work pre-4.10.x
(specifically, I couldn't get 3.x
working). However, I just tried using upgraded CentOS 7 stock kernel (non-mainline), and get the same problem.