Hi guys,

I'm a SAH, SAH Beta and EAH user (inactive) and this is repost of an issue I had with 5.8.3 while using it with SAH. I forwarded a copy of this to Dr Anderson's email but, this maybe a more public (and appropriate?) venue.

Regards,
Wander

-- Message 505678
-- Posted 20 Jan 2007 1:15:40 UTC


I ran into some problems with this version while using it with SAH. Initially I downloaded and installed v5.8.0 to test the CPU usage limit parameter, to put back an old P3 that was suffering with overheating problems back into the game.

After a week or so of testing, I heard that v5.8.3 was available and decided to give it a try, and the problems started:

- although there was no message anywhere (stderr, stdout, Messages, etc) BOINC mgr was working in EDF mode, even though it was not overcommited or otherwise late to report;
- after it finished processing a WU the task list stated the correct CPU usage time, but always something like "1.832%" in the completion column;
- once reported back to servers, it claimed a incompatible number of credits (e.g. 10,000sec and 0.5 credits claimed), some examples: 458271218 and 458318178;
- there was a problem with another WU but this seems to be a different case. BOINC Mgr was unable to finish the UL because it expected the result file to be bigger than it actually was. I was forced to abort the UL;

I suspended both processing and network activities and downgraded back v5.8.0 and all problems seems to have subsided, but when the v5.8.0 first came online, it reported that the state file damaged. To ensure that I wasn't going to trash any other WU, I only allowed 1 WU to report back after the downgrade and waited to see if it would be validated. Now that it has, I opened the floodgates back again.

Although the WUs claimed very little credit, they were validated and thanks to the quorum, they were granted the correct amount of credits :) but they could have disrupted it just as well if the results were turned in earlier.

I know the version is supposed to be for development only, but since this problem occurred in two different machines I'd like to know if anybody else experienced something like that and if I should inform BOINC alpha about this.

Regards,
Wander

Edit: for the record, I'm using Chicken's optimized app in both machines

I saw your post on SAH. Maybe you want to test with EAH as well (reactivate).
As else it's A Seti problem. In my opinion anyway.

---

I saw your post on SAH. Maybe you want to test with EAH as well (reactivate).
As else it's A Seti problem. In my opinion anyway.

Well, I downgraded my machines back to 5.8.0 and although I'm aware of the importance to track down this problem, I just saw a post this SAH Beta describing the same problem I had, where a faulty result was reported by an 583 client, but since it was the second in the quorum they awarded the entire WU something like 0.05 credits, while the other machines claimed something like 15 credits.

As Ageless first suggested and by Dr Anderson's request, I made another test using v583, but this time I used the stock SETI app, and let 4 WUs to be processed:

These two were already being processed by Chicken's app when I upgraded back to 583 and replaced it with the stock SETI app:
460763734
460843516

These were entirely processed by the stock SETI app on BOINC v583:
460845310
460925608

As you can see, all results were correctly processed, uploaded, reported in and validated. Although none of these 'tests' (or previous ones for that matter) can be considered conclusive, there maybe some kind of problem between Chicken's app and v583, and its important to note that a similar problem was reported with v583 and the SAH Beta app.

I hope you find this useful.

Regards,
Wander