CPDN News and Announcements

This forum provides an FAQ for users as well as allowing you to ask questions about MundayWeb's very popular BOINC stats services.
Post Reply
mo.v
Member
Posts: 99
Joined: Mon Feb 19, 2007 7:31 pm
Location: Portsmouth UK

Post by mo.v »

All CPDN servers are now fully functional.
mo.v
Member
Posts: 99
Joined: Mon Feb 19, 2007 7:31 pm
Location: Portsmouth UK

Post by mo.v »

The CPDN upload server climateapps3 has filled up and Milo is moving data from it overnight. Until Monday 3 Dec some CPDN crunchers will find that their trickles and zip file uploads are refused. If this happens to your CPDN model you should if possible suspend network activity for at least 12 hours.

The server status of all ClimatePrediction projects can be checked here:

CPDN
http://climateapps2.oucs.ox.ac.uk/cpdnb ... status.php
BBC
http://bbc.cpdn.org/server_status.php
Beta
http://climateapps1.oucs.ox.ac.uk/beta/ ... status.php
SAP
http://attribution.cpdn.org/server_status.php
mo.v
Member
Posts: 99
Joined: Mon Feb 19, 2007 7:31 pm
Location: Portsmouth UK

Post by mo.v »

The servers are now all functional again.

BBC and CPDN members together have now crunched over 22000 Hadcm (coupled) models.

The Seasonal Attribution Project has stopped handing out new SAP models; as always, crunchers are asked to please complete models they already have. It is hoped that similar SAP-type models will fairly soon be made available as an extra choice on the main CPDN project. Their aim will be to investigate particular climatic events in other years in other regions.
mo.v
Member
Posts: 99
Joined: Mon Feb 19, 2007 7:31 pm
Location: Portsmouth UK

Post by mo.v »

There was a full unplanned CPDN server outage on Wed Dec 5, which Tolu has now put right. CPDN credits will be delayed for a day on the stats sites.
mo.v
Member
Posts: 99
Joined: Mon Feb 19, 2007 7:31 pm
Location: Portsmouth UK

Post by mo.v »

Tolu has added to the CPDN work queue a very small number of HADAM 1-year seasonal models like those previously only available on the SAP project. They take less time to run than HADSM slabs, perhaps 10 days running 24/7. HADAM models should only be run on computers with at least 1Gb RAM, or 1.5Gb for two models run side-by-side on a dual-core.

We expect more of these HADAM models to be available in the new year.

REMINDER: CPDN members can select the type of model they prefer in their project preferences. Availability of models can be seen on the server status page, to which there's a link further up this thread.
mo.v
Member
Posts: 99
Joined: Mon Feb 19, 2007 7:31 pm
Location: Portsmouth UK

Post by mo.v »

There have been two recent problems on the CPDN servers while Milo has been extracting large amounts of model data for the researchers.

* When trying to upload trickles, the message that there's an HTTP error has occurred. It doesn't mean that the CPDN server is down, or that your internet connection has failed. The best idea is to suspend network activity in BOINC manager and allow it again a short time later.

* There have been delays with credits. Don't worry - we will all receive our full credits eventually even if they're sometimes late!

These problems may or may not continue during the holiday period.
mo.v
Member
Posts: 99
Joined: Mon Feb 19, 2007 7:31 pm
Location: Portsmouth UK

Post by mo.v »

The SAP servers are down. Milo has been informed but as it's the weekend he may not be able to investigate the problem until Monday. If you don't want to suspend BOINC network activity and have a SAP model close to completion, you could suspend the model and let the computer crunch tasks from other projects until the server problem is resolved.

The CPDN and BBC servers are functioning normally.
mo.v
Member
Posts: 99
Joined: Mon Feb 19, 2007 7:31 pm
Location: Portsmouth UK

Post by mo.v »

The SAP servers are all up and running again thanks to Milo.
mo.v
Member
Posts: 99
Joined: Mon Feb 19, 2007 7:31 pm
Location: Portsmouth UK

Post by mo.v »

CPDN server problems nearly always seem to happen on a Friday. The CPDN upload servers climateapps3 and uploadatm are down. This will affect trickles and zip file uploads of some CPDN members.

If you only crunch CPDN and encounter this problem, which will show up in BOINC manager messages or by a 10-year zip file unable to upload from the Transfers tab of BOINC manager, the best idea is to suspend network activity using the BOINC manager Activity menu.

For multi-project crunchers who don't want to suspend network activity, the best advice is to prevent your CPDN model producing a 10-year zip file. Set CPDN to No New Tasks and suspend your model in the Tasks tab before it reaches December of a year ending in 0 eg 1970, 2030.

We may have to wait until Monday for this server problem to be fixed. CPDN server status can be checked here:

http://climateapps2.oucs.ox.ac.uk/cpdnb ... status.php

This server problem does not affect CPDN beta, SAP or BBC members.
mo.v
Member
Posts: 99
Joined: Mon Feb 19, 2007 7:31 pm
Location: Portsmouth UK

Post by mo.v »

Milo has just got the upload server climateapps3 going again. The problem occurred because the disk was nearly full.
mo.v
Member
Posts: 99
Joined: Mon Feb 19, 2007 7:31 pm
Location: Portsmouth UK

Post by mo.v »

A few hours ago (Saturday) the CPDN upload server uploadatm was down temporarily but it's now up again.


However, the CPDN servers have been problematic for over a month and this situation may continue for some time. Even when all the CPDN servers are up and show as normal on the CPDN server status page, they have been under stress because of large amounts of data extraction for the researchers and because the disks contain so much data. There may be other undiagnosed problem(s). Tolu and Milo have various plans for the CPDN servers but cannot implement them all yet.

As a result of the current situation

* The CPDN-BOINC forum is sometimes difficult to access and we see a 'too many connections' message. Try later or use the independent forum instead.

* Sometimes the servers temporarily refuse to accept trickles and zip file uploads. BOINC has a rich variety of descriptions for this problem. You may see messages such as

HTTP error
HTTP internal server error
Project communication failed
Scheduler request failed: failed sending data to the peer
Project servers may be temporarily down

These messages can appear even when all the servers seem normal on the server status page. The best idea is to suspend BOINC network activity if possible and try again later.

There have also been delays recently with the appearance of our successful trickles on our models' project web pages.

These problems do not affect the progress of our models and we are all receiving correct credits. Only the CPDN servers are affected, not BBC or SAP.

Many thanks to all CPDN crunchers for your patience!

Edit 20 Jan. It's possible that sometimes a CPDN server shows on the server status page as down when in fact it's running! So in future I will only announce server outages when we have evidence that they're real.
mo.v
Member
Posts: 99
Joined: Mon Feb 19, 2007 7:31 pm
Location: Portsmouth UK

Post by mo.v »

Thanks to Jonathan Colman from Washington DC for pointing out this Nature Conservancy article published on 29 January by Dave Connell. It highlights the ClimatePrediction project. The article is being discussed on Digg here.

There's a forum thread about it here.

Thanks also to Xapadoo for setting up a ClimatePrediction fan for members of Facebook. Read here about how Facebook members can join the fan and invite their friends to find out about CPDN.
mo.v
Member
Posts: 99
Joined: Mon Feb 19, 2007 7:31 pm
Location: Portsmouth UK

Post by mo.v »

We've now crunched over 200,000 HADSM models including Classic slabs and several types on the main CPDN project. We've processed over 10,000 HADAM models on the Seasonal Attribution Project and CPDN, and are fast approaching 23,000 HADCM models from BBC and CPDN members. Project totals can be seen on the CPDN index page where the display is now clearer. We can select the type of model we wish to crunch on the project preferences page of our account.

Many thanks to all members for taking such good care of your computers and models. For members not quite certain how to take care of them, Mike has advice and suggestions in the project READMEs here and here.

Several recent presentations by the research team have been added to the publications section of the CPDN website.
mo.v
Member
Posts: 99
Joined: Mon Feb 19, 2007 7:31 pm
Location: Portsmouth UK

Post by mo.v »

Some of the CPDN main project servers are down for maintenance.

http://climateapps2.oucs.ox.ac.uk/cpdnb ... status.php

The CPDN-BOINC forum and our CPDN account and model web pages are not accessible. Trickles and zip file uploads may fail and produce messages like

26/02/2008 00:50:30|climateprediction.net|Message from server: Project is temporarily shut down for maintenance

The best idea is to suspend BOINC network activity until the CPDN servers are up again, though we realise that this is not always possible for multi-project crunchers. The BBC, SAP and Beta2 servers are all functioning normally.
mo.v
Member
Posts: 99
Joined: Mon Feb 19, 2007 7:31 pm
Location: Portsmouth UK

Post by mo.v »

Tolu worked all Monday and a good part of the night on unscheduled server maintenance, mostly moving large volumes of data. All the CPDN servers are again fully functional.

BBC and CPDN crunchers together have now completed over 23,000 HADCM models. Thank you, everyone! The current 80-year HADCMs are 160-year models that didn't complete on the first computer that crunched them. Many are specially selected to be completed on a different computer because they produced interesting results during their first 80 years.
Post Reply