I am using 1.11.2 version of docker in swarm on my rhel 7.2 corporate machines. I am using devicemapper as the storage driver. It behaves “strange” once overall number of images and containers is over ~1000 and operations such as ‘docker pull’ hang indifinetely until I kill them manually. When I decrease the number of containers and images new operations run smoothly. On my logical volume I have enough space so this is not directly space related.
Are there any known limitations of devicemapper on RHEL or on linux in general? Have you faced some issues with large number of images and/or containers in case of devicemapper’s usage?
In my experience devicemapper has been very flaky especially when the number of containers and images gets high. I’d make sure you are following the guidelines at https://docs.docker.com/engine/userguide/storagedriver/device-mapper-driver/#/configure-docker-with-devicemapper and be realistic about the tradeoffs you are making when using this storage driver. It does not really have a great reputation in general although I have not tried it for probably a year plus at this point.
@nathanleclaire Is that also true with devicemapper+LVM? I know devicemapper+loopback-file has been an easy-if-disrecommended option for a while, but (the last time I looked at this) using a dedicated LVM device was supposed to be better.
(I’m having hard-to-describe trouble, along the lines of “every Docker command, even docker info, is really really slow”, under situations that are roughly “our Docker-intensive automated build system decides it’s going to a bunch of stuff all together”, on the Amazon ECS image. That’s using devicemapper+LVM, but if there’s a better setup that would address the issue, we’d change.)