Hi Everyone,

The MilkyWay@home server will continue to experience maintenance outages today as we work on resolving the problem with the transitioner. Again, thanks for all your patience. You can check the status of the server at any time at https://milkyway.cs.rpi.edu/milkyway/server_status.php.

Best,
Tom

Tom,

If I look at my various tasks pages on the web site they still show the extremely large work-unit numbers (which, of course, can no longer be found!) Am I correct in assuming that the transitioner problem is related to this??? If not, you've got another problem, I fear...

Good luck finishing up the fixes - I know what a pain doing primary-key changes can be on databases of any size (been there, done that.)

Cheers - Al.

P.S. On large numbers of jobs... Every so often, WCG take their databases off-line, do whatever is necessary to avoid referential integrity problems, then subtract a fixed, large, number from all their current task and/or work-unit numbers as necessary before resuming service - I presume that's what you've been doing here! As one shouldn't have lots of very old (low-numbered) tasks/work-units lying around it's a possible "regular" solution without increasing database field sizes.

I've just downloaded and am processing Milkyway@home tasks again.

---

The problem seems to be fixed and the server is running (as far as we can tell) normally.

The issue was actually only tangentially related to running out of valid workunit IDs; a stale run went and managed to generate a task with ID +1 over the INT limit, which couldn't enter the database and was clogging up things. Once we had cleared that out, reset the workunits that Boinc wasn't able to clear on its own, and worked through the backlog of stuck workunits, the server was able to function as intended.

Tom

Working for me with no apparent problems.

---

Plus SETI Classic = 21,082 WUs

Tom tnx for your help / work / fix in the weekend

best regards Wishy

I've had a bunch of tasks, usually de_nbody tasks that should take about 9 minutes, but the longer they run, the longer they have left to run, until eventually they never finish with a "computation error"
DaveH52