Red Hat Bugzilla – Bug 1476310
Downloading image using "openstack image save" fails with [Errno 32] Corrupt image download.
Last modified: 2018-02-24 21:51:12 EST
Created attachment 1305985 [details]
My notes on root causing the issue
Description of problem:
Downloading the image from glance using "openstack image save" fails.
[root@overcloud-controller-0 ~]# openstack image save --file ./example-test.img b5bad39e-4c1b-4d89-8864-0aba2c2be1b9
[Errno 32] Corrupt image download. Checksum was d41d8cd98f00b204e9800998ecf8427e expected ee1eca47dc88f4879d8a229cc70a07c6
[root@overcloud-controller-0 ~]# ls -lrt
-rw-------. 1 root root 1883 Jun 15 17:38 original-ks.cfg
-rw-------. 1 root root 2620 Jun 15 17:38 anaconda-ks.cfg
-rw-r--r--. 1 root root 0 Jul 28 01:44 example-test.img
[root@overcloud-controller-0 ~]# md5sum example-test.img
Version-Release number of selected component (if applicable):
This is seen in OSP 10
[stack@undercloud ~]$ sudo yum info openstack-tripleo-heat-templates
Loaded plugins: search-disabled-repos
Name : openstack-tripleo-heat-templates
Arch : noarch
Version : 5.2.0
Release : 21.el7ost
Size : 1.4 M
Repo : installed
From repo : rhel-7-server-openstack-10-rpms
Summary : Heat templates for TripleO
URL : https://wiki.openstack.org/wiki/TripleO
License : ASL 2.0
Description : OpenStack TripleO Heat Templates is a collection of templates and tools for
: building Heat Templates to do deployments of OpenStack.
After spending some time on this, I found https://bugzilla.redhat.com/show_bug.cgi?id=1391082 that might be related to the issues that I'm seeing.
This is very reproducible in my opinion.
Steps to Reproduce:
1. Deploy OSP 10 with 1 x controller and 1 x compute
2. Scale by adding a compute node to the deployment. This upgrade will fail.
3. Delete the stack, and redeploy with 1 x controller and 2 x compute nodes
4. Upload a cirros image to glance.
5. Now try and Download the image using "openstack image save". You will see that the image isn't downloaded and a file of size 0 is created.
I initially noticed this issue when the I tried to boot an instance using the cirros image in glance. In this case, the instance boot failed after a long time.
Tested this on Newton, couldn't reproduce :(
mshetty, was the image uploaded successfully?
Is it possible for you to share glance logs?