hm, that's interesting to read.
My box with which I crunch MCM does not need manual intervention.
The three other computers with which I crunch OPNG need massive manual intervention.

It seems to me that they are re-sending "old" OPNG tasks from last year
Why do I think so: at the very beginning of OPNG, the tasks yielded credit points between 200 and 400; lateron, this was changed to a figure between ~800 and ~1000.
Since last night, the tasks again only show some 200-400 points. So my suspicion is that just to send tasks out, they use old ones that had been sent out last year already.
So I am questioning how much sense this makes. Just to feed us we get old stuff which had been processed long time ago?

@erich56

When OPNG came back on stream, the work units had a new target (receptor) whilst OPN1 was still working on the same targets as before the hiatus. The sudden change suggests that there was a fairly urgent need to see the results for this target. This new target seems to have required more iterations per docking, with corresponding increases in credit awarded per task.

Now it appears that the "rush job" for the new target has finished (or, at least, paused), so OPNG is back processing the same target it was working on before the hiatus! The task names will have lower batch numbers because they are part of a substantial set of batches of available data that was already in the pipeline, though they are higher than pre-hiatus.

The highest OPNG batch numbers I recorded pre-hiatus were around the 0149280 region, and these new batches for the same receptor seem to be above 0150200; there are huge numbers of possible ligands to try to dock, so it is not too surprising if there is a backlog to clear now the "rush job" seems to be over. And there's no benefit in sending out unwanted work, so unless WCG inform us otherwise I'd tend to assume it's "live" data. It's just tough luck that the dockings seem to be a lot easier to achieve, so lower credit :-)

And in the meantime, OPN1 is chugging along looking at the same target that OPNG has now gone back to.

Cheers - Al.

[Edited to point out that I had not seen wildhagen's post when I posted this. It doesn't alter what I've said, as I tend to believe folks until they are proven really untrustworthy.]

After about 1 day's outage, WCG is back; however, the network problems are even worse; whereas the problems so far were only with downloads, now uploads are also affected strongly.
So all in all: it gets worse instead of better. I am questioning how come, after all these months they took for the transition from IBM to Krembil.

A moment's consideration might explain why difficulties with uploading would probably appear when the system came back from a lengthy period off-line with an active user base out there. I suspect a lot of us expected that to happen - been there before, had the experience elsewhere...

Think of all those client systems waiting to upload a whole day's worth of work, putting in requests at a much higher rate than would normally be expected... And the same magnification effect applies to downloads as not only were there lots of systems out there that got cut off whilst downloading but there would now be lots more "empty" systems wanting a top-up once they had finished uploading... It looks bad now, but start saying it has got worse if it's still as bad in 24 hours time :-)

The upload/download speeds during the first couple of hours after the system came back gave a good indication of the sheer volume of traffic there must have been -- I normally see upload and download speeds 10 times better than I was getting at 08:00 UTC today! And now I've got my normal half a day's worth of work things seem to have settled down as far as I'm concerned (especially as I've turned off OPN1/OPNG for now!...)

It probably isn't realistic to expect any BOINC system to add capacity just to handle the aftermath of major unplanned outages (though some folks may think otherwise), so when a BOINC site has an "accident" the end users will see upload/download problems; I've experienced this at several other sites, and I seem to recall one or two occasions when the old WCG had issues of this type (sometimes without actual down-time!)...

Cheers - Al.

P.S. It's probably a blessing that the BOINC client won't request work if the upload queue is too big, else there would have been even more initial traffic!