Flush request to database

[lipoja]

Flush request to database to prevent UniqueViolation due to parallel requests processing.

[CLOUDDST-20061]

[yashvardhannanavati]

Flush request to database to prevent UniqueViolation due to parallel requests processing.

I wonder how parallel requests processing is causing the UniqueViolation. Although the pullspec is same, the primary key i.e. ID for the two parallel requests would be different, no?

[lipoja]

Flush request to database to prevent UniqueViolation due to parallel requests processing.

I wonder how parallel requests processing is causing the UniqueViolation. Although the pullspec is same, the primary key i.e. ID for the two parallel requests would be different, no?

Bu there is unique on pullspec.
see

iib/iib/web/models.py

Line 233 in 2b5052c

pull_specification: Mapped[str] = db.mapped_column(index=True, unique=True)

Imaging running two requests on two workers almost at the same time.
First worker added this pullspec to db.session.
Second worker checked if pullspec is in DB., it was not there since it was not flushed yet.
First db.sessoin.flush()
Second worker hit the flush few moments later when there was the insert of pullspec again and it failed.

This is my hypotheses, I will sum it up in RCA tomorrow.

The question is if by adding this flush we do not flush some items to DB too early.