In 3.5.0 HA a new MAAS installation has no available architecture for deployments after the images are in synch

[Jacopo Rota]

Install MAAS HA on more than 2 nodes and wait for the images to be in sync. Then on both the CLI and the UI there are failures as no architectures are available.

CLI:

$ maas root machines create hostname=node1 power_type=virsh architecture=amd64/generic mac_addresses=XXXXX power_parameters_power_address=qemu+ssh://ubuntu@10.244.40.1/system zone=zone2 power_parameters_power_id=node1 testing_scripts=none

-> {"architecture": ["'amd64/generic' is not a valid architecture. It should be one of: ''."]}

UI: see screenshot

The same works fine when MAAS is on a single node (no HA)

[Jacopo Rota (r00ta)]

The issue comes from

>>> BootResource.objects.all()
<QuerySet [<BootResource: <BootResource name=grub-efi-signed/uefi, arch=amd64/generic, kflavor=None, base= rtype=0>>, <BootResource: <BootResource name=ubuntu/focal, arch=amd64/hwe-20.04-lowlatency-edge, kflavor=lowlatency, base= rtype=0>>, <BootResource: <BootResource name=grub-efi/uefi, arch=arm64/generic, kflavor=None, base= rtype=0>>, <BootResource: <BootResource name=grub-ieee1275/open-firmware, arch=ppc64el/generic, kflavor=None, base= rtype=0>>, <BootResource: <BootResource name=pxelinux/pxe, arch=i386/generic, kflavor=None, base= rtype=0>>, <BootResource: <BootResource name=ubuntu/focal, arch=amd64/ga-20.04, kflavor=generic, base= rtype=0>>, <BootResource: <BootResource name=ubuntu/focal, arch=amd64/ga-20.04-lowlatency, kflavor=lowlatency, base= rtype=0>>, <BootResource: <BootResource name=ubuntu/focal, arch=amd64/hwe-20.04, kflavor=generic, base= rtype=0>>, <BootResource: <BootResource name=ubuntu/focal, arch=amd64/hwe-20.04-edge, kflavor=generic, base= rtype=0>>, <BootResource: <BootResource name=ubuntu/focal, arch=amd64/hwe-20.04-lowlatency, kflavor=lowlatency, base= rtype=0>>]>

>>> BootResource.objects.get_usable_architectures()
[]

[Jacopo Rota (r00ta)]

All the architectures are excluded here https://github.com/maas/maas/blob/3f1cde4bee61c42f9b07acfcee1708eb1c1db73d/src/maasserver/models/bootresource.py#L592 because apparently

`resource_set.sync_size != (
                n_regions * resource_set.files_size
            )`

[Jacopo Rota (r00ta)]

The issue is here

            resource_sets = self.sets.order_by("-id").annotate(
                files_count=Count("files__id", distinct=True),
                files_size=Sum("files__size"),
                sync_size=Sum("files__bootresourcefilesync__size"),
            )

as the query is FULL OUTER JOINING bootresourcefilesync and the `files_size` are duplicated on multiple rows depending on the number of bootresourcefilesync records.

>>> print(b.sets.order_by("-id").annotate(files_count=Count("files__id", distinct=True),files_size=Sum("files__size"),sync_size=Sum("files__bootresourcefilesync__size"),).query)
SELECT "maasserver_bootresourceset"."id", "maasserver_bootresourceset"."created", "maasserver_bootresourceset"."updated", "maasserver_bootresourceset"."resource_id", "maasserver_bootresourceset"."version", "maasserver_bootresourceset"."label", COUNT(DISTINCT "maasserver_bootresourcefile"."id") AS "files_count", SUM("maasserver_bootresourcefile"."size") AS "files_size", SUM("maasserver_bootresourcefilesync"."size") AS "sync_size" FROM "maasserver_bootresourceset" LEFT OUTER JOIN "maasserver_bootresourcefile" ON ("maasserver_bootresourceset"."id" = "maasserver_bootresourcefile"."resource_set_id") LEFT OUTER JOIN "maasserver_bootresourcefilesync" ON ("maasserver_bootresourcefile"."id" = "maasserver_bootresourcefilesync"."file_id") WHERE "maasserver_bootresourceset"."resource_id" = 6 GROUP BY "maasserver_bootresourceset"."id" ORDER BY "maasserver_bootresourceset"."id" DESC