Description of problem: When deploying Openstack with Ceph, historically, re-deploying Ceph and re-using OSD drives which have not been wiped has caused problems. Many of the field engineers are aware of this and include a "wipe disk" script that wipes all of the drives on a ceph node other than the root disk. This script relies on the gdisk package to wipe those disks. This packages has been in all previous overcloud images that I am aware of but is now absent from the OSP14 ones. Version-Release number of selected component (if applicable): How reproducible: Steps to Reproduce: 1. Instacl gdisk in overcloud-fill image 2. 3. Actual results: Ceph OSD's not cleaned up prior to installing Ceph and the TripleO deployment succeeds but Ceph comes up with 0 OSDs. Expected results: Disks are wiped prior to ceph install. Additional info:
Darin, curious if clean_nodes = true in undercloud.conf parameter would satisfy the need to wipe metadata and GPT labels from both operating system and designated OSD disks in a deployment? I believe this setting will take care of what you are asking for without the need for a wipe disk script, like what had to be done before OSP12.
We have documented the correct way to wipe the disk. Please see the Deploying an Overcloud with Containerized Red Hat Ceph document [1] section 2.1. The gdisk package is not necessary. [1] https://access.redhat.com/documentation/en-us/red_hat_openstack_platform/13/html/deploying_an_overcloud_with_containerized_red_hat_ceph/
clean_nodes = true does wipe the OSD's and is fine when Ironic doesn't leave the nodes in a "non available" state, which happens in the field. Including the gdisk package (660k) is an east trade off to make things run smoothly in the field. If we're that intent on shrinking the size of the overcloud image, perhaps we could consider removing packages like "git" (22Mb)? I also noticed a number of font related packages which are not needed for the overcloud AFAIK? If they are needed for Horizon, shouldn't they be in the container image and not needed at the baremetal layer? Sorry, it just seems like we're cutting useful utilities out of the images when there's plenty of other "fat" to cut.
There is no hard dependency on gdisk for Ceph integration to work with OpenStack, though I agree it is a useful utility, and have no issue with it being included in the overcloud image. If release delivery would consider adding it back the overcloud image, then I agree it would be useful in the field.
There is nothing stopping the field from adding the RPM via virt-customize. The problem here is that there is no operational requirement for this software (and no owner). This wasn't a cutting the fat scenario. It was never explicitly included so once the transitive dependency was removed, so was gdisk. The same goes for the fonts, we aren't explicitly requiring them and they are getting pulled in via some other piece of software that is currently installed (or maybe even included from the base image). It's likely not an openstack package because we removed them all except for the clients in OSP14. I'll take this back and try and get it back in as time allows.
If this bug requires doc text for errata release, please set the 'Doc Type' and provide draft text according to the template in the 'Doc Text' field. The documentation team will review, edit, and approve the text. If this bug does not require doc text, please set the 'requires_doc_text' flag to -.
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHBA-2019:0446