If I suspend network on BOINC all day, when I allow network access again, the uploads begin uploading, the Work Fetcher starts fetching from LHC (but this may just affect the first project in the list). LHC gives me some stuff to download--usually just two results. Once the uploads are done (a second or two later), BOINC Manager automatically hits the LHC scheduler to report what it just uploaded. Which results in the RPC too Recent error and a backoff on more downloads from LHC. Work Fetcher then goes on to hit SETI (project #2 in the list), DL a slew of work, and no backoff (since all the uploads complete while BOINC double-taps the LHC scheduler. After the 58-second backoff from LHC, the local cache is full of Seti results, so no need to grab any LHC work.

I know the auto-connect after upload is by design. Is this behavior the best it can get, architecturally? Or should I just relax, unclench my BOINC-micromanaging nether regions, and let STD and LTD work their magic to make sure that my resource shares are enforced in the long term? I usually leave the network connected, except during business hours (the UL/DL bursts can bork out my IP softphone).

Interesting philosophical dilemma, though. I wonder what this behavior does on modem connections.

(j)
James

I know the auto-connect after upload is by design. Is this behavior the best it can get, architecturally? Or should I just relax, unclench my BOINC-micromanaging nether regions, and let STD and LTD work their magic to make sure that my resource shares are enforced in the long term? I usually leave the network connected, except during business hours (the UL/DL bursts can bork out my IP softphone).

James

Actually there is a bug in the 4.4x versions of the client that causes reporting immediately, it should delay until a trigger event.

I would just let BOINC do it's thing though. It should work itself out eventually.

---

BOINC WIKI

BOINCing since 2002/12/8

Actually there is a bug in the 4.4x versions of the client that causes reporting immediately, it should delay until a trigger event.

I would just let BOINC do it's thing though. It should work itself out eventually.

Aha. I thought that was design behavior. Didn't realize it had already been marked as a bug.

BTW, I'm seeing this on 4.72. Haven't tried 5.1.x yet, since LHC won't support it, but hoping it's fixed in 5.2.

Thx, Keck.

(j)
James

Hmmm, I am not seeing on 4.72 (OS-X) ... interesting ...

I think most of the projects are trying to get ready for the change over. Predictor@Home is supposed to be complient now ... I don't know about SZTAKI or Rosetta@Home yet ...

---

Hmmm, I am not seeing on 4.72 (OS-X) ... interesting ...

I think it has to do with the network being suspended. So when the net comes back up, uploads start. During this time, a trigger event is recognized telling the CC to report results. That's what generates the double tap.

It's a non-standard configuration for me, since I'm on a cable modem, but I'm wondering what the impact will be on the modem users out there.

(j)
James