Bug 1576285

This PR fixes the issue, I could reproduce it locally:

https://github.com/projectatomic/atomic/pull/1235

Atomic host Version:
● ostree://rhel75:rhel-atomic-host/7/x86_64/standard
                   Version: 7.5.2 (2018-06-09 02:40:55)



Tested the bug with following observations:

Scenario-1. 
-------------------
When Container is not stopped and 'atomic uninstall' command is run, the error mentioned is not observed, but the uninstallation isn't clean as it does not create the backup copy at path '/var/lib'. (See output in scenario-2)

[root@ipadocker ~]# docker ps
CONTAINER ID        IMAGE               COMMAND                  CREATED             STATUS              PORTS               NAMES
d21c518ee411        rhel7/ipa-server    "/usr/local/sbin/init"   17 hours ago        Up 17 hours                             ipadocker

[root@ipadocker ~]# docker exec -it ipadocker kinit admin
Password for admin: 

[root@ipadocker ~]# docker exec -it ipadocker rpm -q ipa-server
ipa-server-4.5.4-10.el7_5.2.x86_64

[root@ipadocker ~]# atomic uninstall --name ipadocker rhel7/ipa-server
docker stop ipadocker
ipadocker

[root@ipadocker ~]# ls -l /var/lib/ipadocker/
total 20
-rw-r--r--.  1 root root   37 Jun 11 05:56 build-id
-rw-r--r--.  1 root root   41 Jun 18 06:01 docker-run-opts
drwxr-xr-x. 15 root root 4096 Jun 18 06:12 etc
-rw-r--r--.  1 root root   22 Jun 18 06:01 hostname
drwxr-xr-x.  3 root root   59 Jun 18 06:12 root
drwxr-xr-x.  3 root root   19 Jun 11 05:56 usr
drwxr-xr-x.  6 root root   57 Jun 11 05:56 var
-rw-r--r--.  1 root root    4 Jun 18 06:12 volume-version

[root@ipadocker ~]# docker ps
CONTAINER ID        IMAGE               COMMAND             CREATED             STATUS              PORTS               NAMES


Scenario-2
-----------------
When Container is stopped and then 'atomic uninstall' command is run, the error mentioned is not observed, but the uninstallation process is complete with the backup copy at path '/var/lib'.

[root@ipadocker ~]# docker ps
CONTAINER ID        IMAGE               COMMAND                  CREATED              STATUS              PORTS               NAMES
f6badcd01aba        rhel7/ipa-server    "/usr/local/sbin/init"   About a minute ago   Up About a minute                       ipadocker

[root@ipadocker ~]# docker stop ipadocker
ipadocker

[root@ipadocker ~]# docker ps
CONTAINER ID        IMAGE               COMMAND             CREATED             STATUS              PORTS               NAMES

[root@ipadocker ~]# atomic uninstall --name ipadocker rhel7/ipa-server
docker run --rm --privileged -v /:/host -e HOST=/host -e DATADIR=/var/lib/ipadocker rhel7/ipa-server /bin/uninstall.sh
Moved [/var/lib/ipadocker] aside to [/var/lib/ipadocker.backup.20180619-042114].

[root@ipadocker ~]# ls -l /var/lib/ipadocker.backup.20180619-042114/
total 20
-rw-r--r--.  1 root root   37 Jun 11 05:56 build-id
-rw-r--r--.  1 root root   41 Jun 18 06:01 docker-run-opts
drwxr-xr-x. 15 root root 4096 Jun 18 06:12 etc
-rw-r--r--.  1 root root   22 Jun 18 06:01 hostname
drwxr-xr-x.  3 root root   59 Jun 18 06:12 root
drwxr-xr-x.  3 root root   19 Jun 11 05:56 usr
drwxr-xr-x.  6 root root   57 Jun 11 05:56 var
-rw-r--r--.  1 root root    4 Jun 18 06:12 volume-version


There needs to be consistency in the output / behaviour of the command when run across different scenarios, thus on the basis of above observation changing status to 'ASSIGNED'.

I think the INSTALL label in the image must be fixed for the second scenario to work:

"install": "docker run -ti --rm --privileged -v /:/host -e HOST=/host -e DATADIR=/var/lib/${NAME} -e NAME=${NAME} -e IMAGE=${IMAGE} ${IMAGE} /bin/install.sh"

If there is an `--rm` in the install label then the container is automatically deleted by Docker on a "stop" thus there is no container left after it is stopped.

nevermind my last comment, it is a bug in atomic that I am looking into..

Could you check if https://github.com/projectatomic/atomic/pull/1242 fixes the issue you have reported?

I tested this on atomic host installing latest atomic rpm, this bug seems not fixed, I got the same result as Nikhil had in Comment 7.
Versions:
[root@atomic-00 cloud-user]# rpm -q atomic
atomic-1.22.1-25.git5a342e3.el7.x86_64

root@atomic-00 cloud-user]# atomic host status
State: idle; auto updates disabled
Deployments:
● ostree://rhel-atomic-host-ostree:rhel-atomic-host/7/x86_64/standard
                   Version: 7.5.2 (2018-06-09 02:40:55)
                    Commit: db4a302e874cdd9cc9517a63133cfdf05e23cb684faae166b444c74cf7c146e8
                  Unlocked: hotfix

  ostree://rhel-atomic-host-ostree:rhel-atomic-host/7/x86_64/standard
                   Version: 7.5.2 (2018-06-09 02:40:55)
                    Commit: db4a302e874cdd9cc9517a63133cfdf05e23cb684faae166b444c74cf7c146e8



Basically, I follow the guide[1] to set up the environment:

Scenario-1, uninstall without stop the container:
————————————————
1. pull the latest ipa-server image
2. atomic install --name server-container rhel7/ipa-server publish --hostname $hostname ipa-server-install --ip-address 2001:DB8::1111
3. atomic run --name server-container rhel7/ipa-server
4. On another terminal: 
    1. docker exec -it server-container kinit admin
    2. atomic uninstall --name server-container rhel7/ipa-server

Expect:
The container can be stopped, and the backup copy should be created.

Actual result:
If I execute the uninstall command once, the container can be stopped but the backup copy can NOT be created. But if I execute it again, the backup copy can be created:
[root@atomic-00 cloud-user]# atomic uninstall --name server-container rhel7/ipa-server
docker stop server-container
server-container

[root@atomic-00 cloud-user]# atomic uninstall --name server-container rhel7/ipa-server
docker run --rm --privileged -v /:/host -e HOST=/host -e DATADIR=/var/lib/server-container rhel7/ipa-server /bin/uninstall.sh
Moved [/var/lib/server-container] aside to [/var/lib/server-container.backup.20180912-132434].

Scenario-2, uninstall after stop the container:
———————————————————
Uninstall command can successfully stop the container and create the backup copy.
[root@atomic-00 cloud-user]# atomic uninstall --name server-container rhel7/ipa-server
docker run --rm --privileged -v /:/host -e HOST=/host -e DATADIR=/var/lib/server-container rhel7/ipa-server /bin/uninstall.sh
Moved [/var/lib/server-container] aside to [/var/lib/server-container.backup.20180912-140031].


[1] https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/7/html-single/using_containerized_identity_management_services/index#using-the-ipa-server-container

I think the patch didn't get into a build yet.  I've tried to reproduce locally but I see the uninstall label is honored with the latest version from git:

$ sudo ./atomic uninstall --name ipa-server registry.access.redhat.com/rhel7/ipa-server 
docker stop ipa-server
ipa-server
docker run --rm --privileged -v /:/host -e HOST=/host -e DATADIR=/var/lib/ipa-server registry.access.redhat.com/rhel7/ipa-server /bin/uninstall.sh
Moved [/var/lib/ipa-server] aside to [/var/lib/ipa-server.backup.20180918-203618].
Unable to find 1650979e359c2398d43a03b8c3b053c2e09a9a75387d51bd6530040fa16a9d3f in the installed containers

Would it be possible for you to try with the git version?

I'm afraid the patch is still not going into the build?
I do the same test for atomic-1.22.1-26.gitb507039.el7.x86_64 as Comment 15, and I have the same result too. It doesn't work.
Then I try the git version as Giuseppe recommended, this issue is not seen. Any ideas, Fero?

Hi,

the patch says:
diff --git a/Atomic/backends/_docker.py b/Atomic/backends/_docker.py
index 6030078..56d7a80 100644
--- a/Atomic/backends/_docker.py
+++ b/Atomic/backends/_docker.py
@@ -455,7 +455,6 @@ class DockerBackend(Backend):
             # have deleted it
             if self.has_container(name):
                 self.delete_container(con_obj.id)
-            return
 
         if args.force:
             self.delete_containers_by_image(iobject, force=True)



Looking at the code in the built package atomic-1.22.1-26.gitb507039.el7.x86_64.rpm, the little "return" statement is indeed missing and thus the patch is applied and the issue is supposed to be fixed. Maybe it is time to admit that the fix may not fix the supposed-to-be-fixed fixable issue and that we have a problem somewhere.

First idea: Please check the code you are running whether it has the patch applied. It is in /usr/lib/python2.7/site_packages/Atomic/backends/_docker.py

Second idea: The rpm does not work but the git version does, what exactly is the difference between them? Maybe we have a packaging problem?

Yes, the patch is applied. I change a server to test this again. This issue is not seen either for the atomic rpm or github source code. Probably there was some files I messed up on the first machine I had. Sorry about this. I make it verified now.