We can go to Lifemapper, SOB or ChessBrain.
We can go to Lifemapper, SOB or ChessBrain.
I agree, we should change to something that is usefull. What does everyone think about this?
Predictor is coming back up this PM (PST) without web page, but with work units.....
care to give it a 'gentle try' and ease back into it...
I will likely be there all weekend helping fix it. (Seti hurt us bad) Many memory leaks in code were found... I will continue to help find leaks, and fix Linux kernel and Apache compilation.
If I am there..... I will post here and let you all know...
Chuck
BC
This may sound stupid but why was Seti code use anyways as Pedictor is the third Bionc project. I thought thay would of learned from when Astropulse crashed and then Seti not to use the same code or setup. :?: :roll:
Seti / AP got rolled together... it was decided to push Seti to foreground and AP to background. but they still didn't fix it all..... hence why BOINC core is now at version 4.01 already.Originally Posted by nickth
The biggest issue is Zombie procsses... Sun ignores child termination messages, SysV does not..... None of the BOINC code has termination message code handlers in it AT ALL.... Our task for the weekend will be to write in the new code and put the #ifdef SYSV // #endif around it all.
Seti/AP uses SUN (BSD).... Predictor uses Linux.... (SysV)... the interaction with standard Apache is where the Zombie is created.
The other memory leaks (just plain bad code) are being fixed as fast as lint and other methods can find them.
the architecture is good, just the coding is poor quality. As for the DB... Predictor knows better and does not use the same DB architecture / indexing as Seti... so that part is already corrected (and has been since day 1)
Predictor is down at the moment... but by the time I get there, it will be up (or shortly thereafter)... I will let you all know...
Please just ease into it as *I* am going to nurse the servers along and will be a 1-man show.
BC
Originally Posted by em99010pepe
Think I'll run mostly SOB while I wait for things to get straightened out.Originally Posted by Keith75
Since I'm on the P@H team and fixing the Kernel now (found yet another flaw... bad kernel config by the guys who built the 'bigmem' kernel in their 'IT department' (forgot the 'SH' in front of that))....
I will stay on P@H and get the first couple downloads and make sure that I have mem flushing working correctly.
Here is the exact situ..
it COMPLETELY fills 4GB of ram, (just generating WUs) because the buff and cache fill ****THEN ****** starts flushing memory, which takes some 30 seconds... causing swapping all the way.... compounding the poor performance.....
****SOMEONE***** does not know how to gen a Linux kernel.
I'm retuning the kernel and going to set some reasonable parameters... like flush memory every 15-30 seconds, or when min-free = 256M... stuff like that... This default 'bigmem' kernel config is apparantly never meant to swap.
I will test it out... I may ask for a few volunteers in a while, but I will do so in the Predictor thread..... until then, I'm going to give Beerknurd a hard time on D2OL and just catch up to him.... :P
BC
Chuck,
I'm at home all day today, If you want to use me as a guinea pig, just say so.
It's 10am Saturday, the German F1 Grand Prix qualifying start in 2 hours, so I don't leave the house.
Does make you wonder how these so programmers make into the real world.
Stu
Don't forget about MOTOGP. British Grand Prix at Donington Park.Originally Posted by Ototero