DataIntegrityViolationException on HFJ_RES_TAG

[rbhman]

Hi all,

I'm experiencing an intermittent problem with my multi-threaded client inserts, possibly related to first-time multi-threaded writes related to a given resource type,  where the resources have identical profile values.

I have a unit test that writes four medication statements to the JPAServer using GenericClient from four individual threads. Each of the medication statements uses a meta profile which I believe results in a write to the HFJ_RES_TAG table for said profile value.

The very first time I run this test one thread will be successful and the other three will fail with:

ca.uhn.fhir.rest.server.exceptions.InternalErrorException: HTTP 500 : Failed to call access method: org.springframework.dao.DataIntegrityViolationException: could not execute batch; SQL [insert into HFJ_RES_TAG (PARTITION_DATE, PARTITION_ID, TAG_ID, RES_ID, RES_TYPE, PID) values (?, ?, ?, ?, ?, ?)]; constraint [fkbfcjbaftmiwr3rxkwsy23vneo]; nested exception is org.hibernate.exception.ConstraintViolationException: could not execute batch

The constraint listed is for: 

 alter table HFJ_RES_TAG 

       add constraint FKbfcjbaftmiwr3rxkwsy23vneo 

       foreign key (TAG_ID) 

       references HFJ_TAG_DEF (TAG_ID);

So it appears the other three threads failed - because the TAG_ID is not yet committed? 

Following this test, a row for the profile will exist in HFJ_TAG_DEF, and a single row will exist in the HFJ_RES_TAG table for the one successfult thread.

If I run the test again, it will succeed for all 4 threads - every time. I can then reliably re-create the problem by manually deleting the HFJ_TAG_DEF (after first deleting the HFJ_HISTORY_TAG reference) and the related HFJ_RES_TAG rows.

The upshot of this is that the first time I run mass updates of a given resource (e.g. in a new environment), many threads will fail and I have to run everything again, sometimes several times as the error count decreases.

I guess I could run some sort of limited sub-set to try to seed the values, but I would prefer not to have to try to identify what this would be.

Would this be considered a bug, or just something I have to live with in my multi-threaded set-up?

  

[rbhman]

I neglected to add this error output from JPAServer.

Caused by: org.postgresql.util.PSQLException: ERROR: insert or update on table "hfj_res_tag" violates foreign key constraint "fkbfcjbaftmiwr3rxkwsy23vneo"

  Detail: Key (tag_id)=(318) is not present in table "hfj_tag_def".

The tag_id 318 is the one (successfully) created by the first thread. I guess the other threads are attempting to reference that tag but it does not yet exist in their JPA memory model?

[James Agnew]

Oddly enough, this bug got fixed on the weekend.

Current snapshot builds will have the fix, and the upcoming GA release this month (5.4.0) will also have it.

Cheers,

James

--
You received this message because you are subscribed to the Google Groups "HAPI FHIR" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hapi-fhir+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/hapi-fhir/6ad97487-cd6f-469b-89dc-ee386ec51c6bn%40googlegroups.com.