Before this weekly outage more than 900K of WU's were in cache ready to send.

After, the cache is empty and is now slowly filling with new splits.

I have a question for the observant users ;)

I know that the cache contained data from the tapes that was split last.

These 900K WU's deleted must have been considered to contain errors and therefore pruned.

Are these tapes on the present splitter list?

I did not note the names of the tapes that were split during last week, but maybe someone else did?

ChrisD

Matt posted in the latest Panic Mode On thread here.

The thing we tried last week (science database updates in advance of Green Bank data splitting) that didn't quite work? Well, we're doing it again this week, this time with hopefully more success. The beta project and other science database related stuff is offline until this is finished. We may likely run out of workunits, but we shall see...

- Matt

Cheers.

Considering that this week there was another extended outage thus individual caches were more drained than usual it is hardly surprising that the RTS cache on the server was emptied almost as soon the servers came back. The RTS will get back to whatever version of "normal" is "normal", and then it will be time for next weeks' outrage.....

(Perhaps all this activity in getting the servers ready for data from the GBT is indicative that we are not very from the release of its data into the wilds of SETILand....)

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

With a RTS of 900k they have about a 9 hour cache when the return rate has been about 100k/hour.

The last maintenance was about 12 hours, so it's easy to see why all those tasks almost instantly vanished when they came back online.

I would guess it will be 2 more days until there is a substantial RTS, and that depends on if they throw some more AP tasks in the bucket.

---

Brent. I kept a close look at the server status and I did not notice the WU's comming back in the cache. (Only staff can tell if they did.)

For a long time the cache was observed being as good as empty, showing less than 10 W's available. Every request for more work got me only a couple of WU's at a time.

It lasted several Hours before my cache started to fill at a 'normal' rate.

ChrisD

It lasted several Hours before my cache started to fill at a 'normal' rate.

That's absolutely normal, though this week was an extreme case.

It's the old "how long does it take to fill a bathtub" maths problem, with various combinations of taps (faucets) available for filling, and a plughole (us) for emptying. There's an added complexity, in that there's a blockage in the drain below the plughole (us again): the plughole has a high capacity, and drains quickly when it's been dry, but the drain has a more limited capacity, so the flow slows down as the drain fills up.

The splitters are our taps/faucets, and we are the plughole/drain. Every Tuesday, both apertures are closed during maintenance, but there's usually water left in the bath. After maintenance, both are opened again, but our drain is usually faster to get going than the supply taps, so the bath often empties completely.

This week, for operational reasons that Matt has written about, the taps weren't turned back on until the following day, so both the bath and the drainage pipework were both drier than usual. So when the taps were turned back on, everything flowed straight down the plughole: nothing was left to fill the bath. Now, the drain blockage is starting to show again, and water is starting to rise in the bath.

If you wish, you can follow this graphically over various timeframes here.