In recent upgrades we have observed that the module is successfully upgraded (for a tenant) but after further investigation have found errors in log and (in some cases) some issues with database schema.
For example what is observed in the following
I would have expected the migration to have failed (instead of exception being thrown and module upgrade succeeding)
Noting this issue for mod-srs (but would be expected behavior for all modules)
How to reproduce:
- Run mod-source-record-storage-5.0.6
- Install tenant diku.
- Stop mod-source-record-storage-5.0.6
- Start mod-source-record-storage-5.1.5
- Upgrade diku from mod-source-record-storage-5.0.6 to mod-source-record-storage-5.1.5
- Call the API that reports the async upgrade result, for example /_/tenant/458b6397-c4d2-46b2-a771-ace8a310bcea?wait=30000 (for details see https://s3.amazonaws.com/foliodocs/api/raml/r/tenant.html#tenant__operation_id__get and https://github.com/folio-org/okapi/blob/master/doc/guide.md#install-jobs-and-asynchronous-operations )
Note that there is a Liquibase exceptions in the log failing the upgrade but the response has not "error" property.
Update folio-liquibase-util to a version with the fix provided by
LIQUTIL-17 = https://github.com/folio-org/folio-liquibase-util/pull/14
I've tried this and the result is as shown in the "Expected result" section. No change to RMB is needed.