Hi James,
I checked with one of the engineers who work on Eclipselink (and who was
originally helping us debug this problem) and he mentioned that there
was no effort to explicitly tried to analyze this exact problem.
However, he also mentioned that there has been a lot of work done to
improve behavior of EclipseLink under high concurrency.
I recall that we observed this issue on a Niagara platform, and when we
had reported this issue, they were unable to reproduce this issue but it
was not on the same hardware that we had.
I'm not sure what platform you are running your test on, but it may be
worth running your test after reducing the allocation size.
Unfortunately, I cannot test with with the original hardware that we had
when we discovered this issue because we had migrated away some time ago.
Kim
Hi,
There is mention of an Eclipselink bug in the source and the older docs for the
Java app regarding duplicate sequence numbers being generated under heavy load.
Are there any other details regarding this? I didn't find anything in Jira,
looks like it has been in the code since the initial checkin. Has anyone
tested removing the workarounds from the code yet?
thanks,
james
