14 December 2014

14th of December

Farm news
Despite the weather this week has actually had a bit of productivity. While it was hot most of the week we got a couple of cool days and I took the opportunity to fire up the GPUgrid crunchers. I had them running until midday when it started to get hot again.

I also managed to finish off the ClimatePrediction work that one of the Intel-GPU machines had so it has cleared its cache as well. As is usual I got a few file transfer failures for CPDN and GPUgrid which I have to clear via the dial-up.

On Saturday I picked up the two new i7 replacements from the computer shop. Now all I need to do is decommission the couple of old machines and sell them on eBay. The shop still need to supply a copy of Windows 7 for one of them, but that is just a CD and license sticker so I can pick it up during the week. I took a few photos so I will do a separate blog post for them.

BOINC testing
On the BOINC front we got a new build, 7.4.35 which has already been replaced with .36. I was doing some testing of notices with .35 which doesn't seem to notify when it has new notices. The suspend GPU tasks bug has been fixed though.

07 December 2014

7th of December

Farm news
Mostly off would be the best way to describe it. Parallella's and Pi's have been running, everything else has been off. I even had them off for 2 days when it got too hot. We've also been getting a thunderstorm come though pretty much every night this last week with lightening and strong winds.

I did manage to get one of the Intel GPU machines to pick up some ClimatePrediction work and then had it working in bursts of 12 hours (overnight) to try and finish them off. They also give me issues uploading and I have to resort to the old dial-up modem to clear the stuck file transfers. They are taking 52 hours to compute and I restrict them to 4 at a time. I have only just now, managed to complete the first 4.

The replacement 6 core/12 thread machines are still in the works. The shop got the DDR4 memory this week and the first machine is now complete. The second is waiting on Windows to come in from their supplier. I will collect the pair of them next weekend.

I have been selling off some of the old graphics cards on eBay to get some money towards a pair of GTX970's to replace the pair of GTX670's that I use as GPUgrid crunchers. I am working my way from the most recent cards to the oldest ones as I won't get as much money for the old cards. The more recent ones also have their original boxes so it makes it easier to sell and post them.

Raspberry Pi's and Debian
Debian (the flavour of Linux that the Raspberry Pi comes with) are coming up to another release called Jessie. The current release is called Wheezy. I have been running Jessie on all the Pi's for some months now with them getting frequent updates as it was considered a testing release. My main reason for using Jessie was to get a more up to date BOINC client. Jessie went into a freeze period at the beginning of November as they prepare it to become the next release, some time in early 2015 although there is no official release date.

The last BOINC client that made it into Jessie was 7.4.23 so that is the version we will be running for the next 12-24 months, unless you want to compile your own.

BOINC testing
After the release of 7.4.27 they found a couple of bugs and so we're testing 7.4.32 at the moment. The main fix is to the backup project not working (zero resource share). While that seems to be fixed it won't suspend GPU tasks so there will another version to test in due course.

Project news - Seti
They received some data from the Arecibo radio telescope, only 4 disks worth, so that will be processed fairly soon. They are still trying to rebuild their Astropulse database so they haven't managed to get any Astropulse work sent out yet. I understand they are still working on the back-end plumbing for the Green Bank radio telescope data as well.

22 November 2014

22nd of November

Farm news
Weather hot, like in the mid-30's (degrees C) hot for a couple of days this week so the entire farm was off. Its back into the upper-20's so I have the Parallella's and Raspberry Pi's going again. I have been running the Intel-GPU's overnight as well.

The Intel-GPU machines are all doing Asteroids work. Credits are around 14 million so still quite a way to go before it catches up to the other projects. The Parallella's and Pi's are doing Einstein work. I have 2 Parallella's and 6 Raspberry Pi's.

I was in the computer shop this morning dropping off parts for the 2nd new build machine. They are still waiting on memory before they can be completed. It seems DDR4 memory is still in fairly short supply at the moment.

Pi B+ rebuild
A day after I did the last post I went and upgraded the B+ Raspberry Pi and after rebooting it refused to come back. I ended up wasting the next 4 hours restoring from the official image (downloaded back in June) and then having to upgrade it to get it back to what it had and remembering all the things that I normally change from the standard Pi image. I have since taken a snapshot of it so at least I have an image I can just dump onto it should it happen again.

Before anyone asks "why didn't you use one of the other Pi's images" I would point out they are all B models. Their image is a bit different and doesn't work on the B+ which is why I had to go back to the official image and then upgrade it.

Ideally I would like a stripped down image that doesn't have all GUI stuff that wastes space and CPU time. X11 and the desktop stuff are all useless when remotely accessing the Pi. I do remove some of the packages that come with Raspbian but not enough to get rid of x11 entirely.

15 November 2014

15th of November

Farm news
After an almost complete shutdown earlier in the week when the temperatures were in the mid-30's we're now back up and running the Intel GPU machines. When I say "almost complete" the file server and router were off as well as most of the computers. I left the Parallella's and Raspberry Pi's running even though they had no network access.

There is more hot weather forecast for the coming week.

BOINC news
7.4.26 (Mac) and 7.4.27 (Windows) were released to the public. Nothing new for me as I was already running .27 on most of the machines anyway. There have been a couple of fixes done since but we haven't seen a new build yet..

CPDN issues
This morning I picked up a bunch of CPDN work units. They showed up in BOINCtasks as using zero CPU time and were using less memory than some of the other CPDN work units. I left them running for 10 hours. When I checked again they still had the same symptoms so I went and looked on one of the machines and there were Visual Fortran pop-up boxes with error messages. Once I pressed the Okay button the tasks fail. I had 15 of these across 3 machines.

Strangely enough another 5 of them are running and using CPU time, more memory, not to mention doing checkpoints.

Upon reading the CPDN forums for Visual Fortran errors regarding a different climate model, it seems they don't like Windows and Intel iGPU's. Its something to do with the graphics the model can show while running, not that I ever bother with the graphics. It doesn't seem to effect Linux computers.

Seti news
It seems they are scraping around for data to process. They have some older data that needs reprocessing with new apps but that is about all they have from the Arecibo radio telescope. They do have data from the Green Bank telescope but no apps or back-end plumbing (work unit splitters, validators, etc) for it yet. They are also sorting through database issues.

This isn't causing me an issue as I have been concentrating on Asteroids more recently but it seems they now have more processing power than they can handle.