Project

General

Profile

Bug #5410

provenance data is not flushed to HSQL database before Kepler quit.

Added by jianwu jianwu almost 8 years ago. Updated almost 8 years ago.

Status:
Resolved
Priority:
Normal
Assignee:
Category:
provenance
Target version:
Start date:
05/24/2011
Due date:
% Done:

0%

Estimated time:
Bugzilla-Id:
5410

Description

Currently, provenance data writing to HSQL database has write delay (100 ms by default). It causes some data is not update or written because the data is not flushed before Kepler quit.

To reproduce it,

1) Run workflow in Kepler GUI or non-GUI mode.
2) Quit Kepler if Kepler is run in GUI mode right after workflow is executed.
3) Use HSQL client to connect hsql provenance database.
4) Check 'workflow_exec' table: the record the execution might not be written in the database or the execution status might be always 'running'.

It should work for both GUI mode or non-GUI mode. Not sure whether it works for other databases, such as Oracle and mySQL.

History

#1 Updated by Derik Barseghian almost 8 years ago

I'm a little surprised you're seeing this with write_delay at 100ms. Have you tried 0 to see if it's truly the problem? I set write_delay to 100ms at r20931 (see bug#4325) to avoid a similar problem, and that was fast enough. Lowering it is probably fine, but keep an eye out for performance issues - e.g. does workflow execution speed slow down as it waits for writes to provenance?

#2 Updated by jianwu jianwu almost 8 years ago

I discussed this issue with Dan. He confirmed the bug. My test is done in non-GUI mode, so Kepler exit right after workflow execution is done.

I know set write delay to be 0 will slow down workflow execution, maybe quite a lot. So a better way is to keep some write delay and flush data before Kepler exit.

#3 Updated by Daniel Crawl almost 8 years ago

Fixed in r27741.

#4 Updated by Redmine Admin almost 6 years ago

Original Bugzilla ID was 5410

Also available in: Atom PDF