PDA

View Full Version : Site down?



Mustard
08-14-2004, 02:53 PM
Been trying to get to the predictor site again today, no workie that I see. Anyone else able to reach the web page?

Bok
08-14-2004, 04:56 PM
Looks down to me.

BECL
08-14-2004, 06:23 PM
Sorry,
beeen out.... Just got back, called and found out they are there.

It should be up in a little while....

I asked why play with what works.... and that surge this morning had
sent it spinning.... all they want to do is stop the 'spinning' during a surge.

It gets into an I/O bound state swapping (even w/ 4Gig of ram)... they're
processing off the upload surge and generating more WUs.... There were
over 10000 WUs that got dumped by a few all at once.... (this is why
batching is not advisable for any BOINC project.... it takes a LOT of horsepower
for proteins as it is... and BOINC w/ quorum = 3.... takes that much more.

The Apache config has been turned down a few more steps and that should
stop the swap-spinning... since it has a direct relationship on the DB, transitioner, etc.


I will advise if I hear anything different. Also as uptime gets more exact, I will let
you know...


Chuck

PS: I'm on the admin side of the site now and helping...

BECL
08-14-2004, 06:35 PM
Edit......... We're on the way back up.... please go slow with it...
I just got some work... so everyone should get work here shortly.


everyone is trying to hit it at once.


In the future, when you see 'Master File Fetch' failure... you KNOW it's in restart.

and just that fast after 6 MFF's... I got work.


*EDIT*: Keep filling the shared memory faster than it can be emptied... WISH WE HAD THOSE OPTERONS NOW!!! *end edit*



Chuck

BECL
08-14-2004, 07:51 PM
Ok,

Here is the official word...... I quote...




chuck,

Unfortunately we had a sever file system crash today morning. I am now recovering the system (I just came back from the lab). Because of the characteristics of the BOINC framework, it will take a while. i hope to be able to put the system back for Monday. I asked XXXXXX to put some news in the web page of BOINC and SETI. I hope he had a chance to read my message.

I will keep you updated.

Best wishes,
XXXXXX




So, unfortunately until the filesystem (aka DB) recovers from the surge, we are down. I will keep an eye on it and when it makes it back, I will let you all know.


Thanks for your patience and understanding... This is not easy for either us the users OR us on the support staff.


*Edit*

I would recommend going back to your other project(s) until I've had a chance to shake this out a bit and I will notify the team when it's back up.... Knowing how hard they are working on it there..... It will most likely be back up well before Monday

*END EDIT*



Chuck

Spore
08-14-2004, 08:01 PM
That is what I meant - It goes down almost everyday!!

BECL
08-14-2004, 08:11 PM
Originally posted by Spore
That is what I meant - It goes down almost everyday!!



Then, given your lack of confidence in the project at this stage of Predictor (it does say "ALPHA" at the top of the web page).

Then my recommendation to any DC'er who wants a more stable project would be to go back to the 'main stream'.

You won't get any objection from me. I understand we each have our goals, desires and, in this case, limits as to how much instability we are willing to tolerate. This is 'Free-DC' afterall, so please... do whatever suits you best.

I'm going to run my other projects and keep an eye on Predictor, but that is my choice. It's what I signed up to do.

Go DC, have fun.

Chuck