SETI@home News Vol. 116 (19/09/2010)

Leave them be. With the DB down, time-outs are only relevant if you're not done by the time the DB comes back (which will presumably be Friday).

As for my manual reports, I'll just shift the whole BOINC directory elsewhere - then I can turn that machine off.
 
They've left the upload server up - which is good and bad. Good, in that at least uploading can continue, bad in that reporting large batches of tasks is problematic! Damn, we seem to be in a loose-loose situation here! :rolleyes:

Talk about the project staff sucking on fire hoses! They've created a beast that they can't take off-line, otherwise everything backs up due to the limited resources - but they need to take it off-line to better distribute those resources! Catch 22 :(
 
They could have course 'fix' the nightmare by letting everything bar the splitters run. That'd clear the pipes, then they bring everything back on with tight work limits so everyone gets a go. The longer back-offs on upcoming clients should help too.

Mind you, the biggest issue right now is clearing out the flaky machines - which I believe are Bruno and Mork. Bruno has all but gone now, but I think it was Mork that brought everything down last night. That's the peril with unsupported and untested pre-production hardware.
 
Last edited:
Certainly, they need a plan to stabilise the hardware. I think doing as you suggest would be the sensible approach - perhaps even bring the splitters into the equation very early on with very tight server managed limits. I think though that they will be too focused on their 3 day window of 'science' (which may in itself be required to justify continued funding) to follow through on such an approach :( Also, such an approach may itself bring other problems in that the frequency of client connections would increase massively (albeit to report and gain less work) - and those that choose not to inform themselves about what is happening (the so called fit and forget brigade) would continue to hammer the servers trying to get more work.....

It still astounds me that they run the forums out of what is effectively their production database; Sunday proved how much that slows things down. I know there has been talk of hosting the forums elsewhere, and I can't help feeling that they need to re-visit this idea. The forums should not be allowed to bring the main db to its knees. Its been too many byears since I worked with server based databases, but I'm sure MYSQL instances can be replicated through the equivalent of snapshots.

They just seem to be trying to fit too much into either of the 2 windows they have, and the same with the hardware and bandwidth! Lets face it, SETI is now officially creaking - very loudly :(:(:(:(
 
Well my rig is gonna be out for a while it's not stable at any speed any more think my CPU has had it power supply works in mates rig with all my GPU's folding away as a test so Meh :( screw it SETI can wait cos I know I have.
 
ToxicTBag said:
Thanks for that P :)

It's a sad day when Five Hamlet quits seti :(
Click to expand...

I have mixed feelings about him.

Well he posted a lot when they started the 3 day outages - and threatened to quit quite a lot then. If he wants to throw his toys out of the cot, thats his choice - they're his toys. Yes it will be sad to see a significant financial/data contributor go, but I'm sorry - I just don't understand people who go off on one like he has. We contribute computational capacity for the good of the project, and we are here to serve the project - not the other way around. Yes its nice to keep machines rolling, but if there are gaps in availability, so what? the sun will still rise tomorrow and who knows, we may just have a couple of trouble free weeks which will enable the cycle to return to normal. I suspect he will be back - perhaps not for a while though.
 
You must log in or register to reply here.
Back
Top Bottom