[PULP-747] Made pulp-import-export work in the presence of domains. #6927
Conversation
|
Cross-domain checks needed. More importantly (and harder) - the multi-table-nature of Exporter/PulpExporter will require significant re-architecting to have cross-domain-uniqueness work. Behavior in this PR is "creating an Exporter or Importer 'consumes' that name across the Pulp instance". Given the real-world usecases for import/export (specifically, this is an instance-administrator-only function) - may be worth accepting this fix "now", and letting untangling that issue be a future piece of work. The current test cases and workflows cover "domain-enabled to domain-enabled" and "no-domain to no-domain" tests. Test cases for the mixtures (domain-up/no-domain-down and no-domain-up/domain-down) require multiple Pulp instance booted with different settings. |
ggainey
force-pushed
the
747_pie_domains
branch
from
51ff2c6 to
8e84f4e
Compare
ggainey
force-pushed
the
747_pie_domains
branch
from
8e84f4e to
a1c30d9
Compare
|
Current state includes doc, enabling FSExporter, and addresses the unique-constraint Fun. |
ggainey
force-pushed
the
747_pie_domains
branch
2 times, most recently
from
cb454d5 to
e1b0d9c
Compare
|
Looks good so far! |
ggainey
force-pushed
the
747_pie_domains
branch
6 times, most recently
from
162df37 to
83223b9
Compare
ggainey
force-pushed
the
747_pie_domains
branch
2 times, most recently
from
d2672de to
59015ab
Compare
| # IF we're domain-enabled: | ||
| # REPLACE "their" domain-id with ours, if there is one. | ||
| # If not, INSERT "our" domain-id into the path in the right place. | ||
| # Otherwise: | ||
| # REMOVE their domain-id if there is one. | ||
| # Do this before letting QMR run, since it will replace "their" domain-id with "ours" |
There was a problem hiding this comment.
Could we remove the domain on export if it exists and then always expect there is none on import?
(I think this will not only produce smaller files, but also produce a bit more consistent results.)
There was a problem hiding this comment.
That just splits the work into "half on export, half on import" - one ends up doing the same amount of work. And "smaller" by removing domain-ids, when we're export artifacts, is lost in the noise.
There was a problem hiding this comment.
This is certainly not a deal breaker.
I just thought information that is not even in the export can never create confusion on import.
And it would be "more consistent" if non-domain-enabled and domain-enabled systems produced the same format.
Added tests for fs-export. closes pulp#6926.
ggainey
force-pushed
the
747_pie_domains
branch
from
59015ab to
f5f236a
Compare
2 participants
No description provided.