July 22, 2004
We found and fixed the no results bug. It was a malformed embedded SQL query. But we are leaving the scheduler off overnight to let the transitioner work through it's backlog. The transitioner works a great deal faster with a small backlog and we hope yesterday's changes will go a long way in preventing a large backlog from recurring.
July 22, 2004
Yesterday we introduced a bunch of performance enhancements to the system. During the course of letting the system run overnight, we have detected a bug that keeps results from being spent to hosts. It looks at though this bug has only been around for about 13 hours. While we are debugging this issue we have decided to shutdown the scheduler and try and let the transitioners pick up the pace a bit.
July 21, 2004
The log files are currently being compressed, after thats done the project should return to normal.
July 19, 2004
We have brought up an additional transitioner to try and cut through the backlog we have accrued. Since the beginning of the project we have already processed 2.7 million results. For comparison, the beta project had only processed 6 million through the whole life of that project to date.
July 19, 2004
Sometime between 5pm and 12am PDT, the splitters seem to have stopped producing new workunits. We have restarted the splitters and they are back to creating new workunits.