8th Annual BOINC Pentathlon: 5th May - 19th May 2017

In my (very limited) experience, I get messages from various projects I've tried upload-blocking, saying "0 tasks downloaded, too many uploads in progress" or something to that effect, so seems one of the many multiple client methods is about the only way.
 
Did you start boinc each time as a completely new client machine, or did you try and retain any part of it's identity so that your client would have been seen as trusted? If you kept the boinc identity others have reported that something is buried in the client and/or server that tells it that you already have WUs downloaded, but of course it can't then find them, gets confused, and you are in a world of hurt.

I started them each time as a completely new client machine, with different computer name. What would happen is the first machine was called, let's say, rig1, it would show up as rig1 at the WCG, but when rebooted into the other OS, and run rig2, rig1 would be renamed rig2.
 
You need to make sure that a different Computer ID is generated, I'd always suggest creating a second one to check for this before filling up the first with work and completing it.
 
In my (very limited) experience, I get messages from various projects I've tried upload-blocking, saying "0 tasks downloaded, too many uploads in progress" or something to that effect, so seems one of the many multiple client methods is about the only way.
My test was partially successful, uploads were blocked, but as you say, it seems Boinc will not download more work until it can upload completed tasks.

Interestingly however, when i disabled the rule and allowed uploads, Boinc's first action was to download the new tasks before uploading the completed. Additionally, the individual tasks seemed to have their own retry counters. So the may be a 5 - 10 second window of opportunity to disable, then enable the block to download new tasks with only a few be uploaded before you can block them again. A bit hands-on perhaps but you may be able to bunker quite a cache. I will try later.
 
WCG is the problematic one. Initially whatever you do you can only get a very few WUs, enough to fill your threads and one or two spares. Once you return the first couple WCG then gives you limited trust and you can then set your queue to a reasonable period and it will let you have a load of WUs. But there is still a limit. You will never get past that Zika limit per client and at that point must return WUs to get more. You won't be able to bunker enough for the entire 5 day lead in (now down to just under 2 days), but anyone bunkering now, if they were to set their boinc to get say 3 days work will fill that queue. Even so your PC at this stage isn't fully trusted so you'll need a wingman to validate those WUs. But with enough returned valid WUs your PC will be fully trusted. Once it is the Zika project appears to then not require a wingman for WUs downloaded after that full trust is granted. That is of course the holy grail for the pentathlon as your going to be certain of points for the team in the limited window of the project.

I'm not trying to refill caches during this initial bunker period. Having set my boinc at the outset to grab say 6 days of work and a reserve, it got the full allocation of zika, and then topped off with cosmology. I can then block the uploads and not worry that I needed to get any more WUs. As of now my PCs are still crunching both zika and cosmology. I might possibly run dry for a couple of hours tomorrow evening, but that's not enough for me to fret about as once the project starts my host file settings will be deleted so I can upload / download with normal size caches.

The firewall is an alternative way to deal with the prevention of uploads, but as explained above, Zika won't give you more than a set number of WUs, you have to return to get fresh ones. But set a decent cache days now, and if you start now you'll probably have enough for the limited bunker period that remains.
 
I've checked the Linux machine-id on each box and they're all unique.
I didn't lose any work , I loaded up the first then went straight to the second one to test, good job i did.
Not sure what you mean by 'machine-id'.
Computer ID is the unique number given out by a project for each client, some projects are strict and will only give 1 Computer ID per computer.
WCG allows separate Computer IDs for each client, so not strict.
Do you still have both the rig1 and rig2 listed as separate Computer IDs in your list of computers at the project (whatever the 'rigX' number might be) ?
The name of a client (eg rig1) in the list of computers does not have to be unique.
 
ye sorry I miss typed at first. I typed WCG but then the last bit was about cosmology lol. when reading his reply he said hes moving onto cosmo but running WCG now and I told him I couldn't see him on the ocuk cosmology page, but that's because he isn't running cosmology yet
 
Not sure what you mean by 'machine-id'.
Computer ID is the unique number given out by a project for each client, some projects are strict and will only give 1 Computer ID per computer.
WCG allows separate Computer IDs for each client, so not strict.

Sorry, I thought you meant the ID generated by the OS. I can see the different ID numbers on "normal" boinc projects but not on WCG? just the device name.

Do you still have both the rig1 and rig2 listed as separate Computer IDs in your list of computers at the project (whatever the 'rigX' number might be) ?
The name of a client (eg rig1) in the list of computers does not have to be unique.

Although I can't see the IDs they have to be the same given to rig1, rig2 & rig3, because they are never listed as separate device names. On the results status page, from the Device Name drop down menu, I can see rig1, if I now reboot into rig2, rig1 gets renamed to rig2, the server thinks they are the same rig and detaches the work units it gave me before (which are on rig1) because it sees them as not being there no more.
 
The Computer ID is also shown when you click on the project's Properties in Boinc Manager, so you could compare there.
Although WCG is not strict it may still be that a function of a later version of Boinc is required for it to work on a computer, I don't know for sure if that is true or at what version it might occur but I believe I have witnessed some computers happy for multiple clients whilst others would fail (for the same project).
 
I have the latest recommended version of Linux 7.2.42, though it is old (Feb 2014) The only newer version is a development one, but again it's old (Sept 14) But now I know what I'm looking for I'll give it a go, but only AFTER I've uploaded, I daren't try it now.
Thanks.

I'm wrong they aren't the lastest Boinc versions, BOINC development doesn't build BOINC anymore and my Ubuntu distro is old and that's the lastest version in the repo, I'll need to upgrade if I want a newer version.
 
Last edited:
Just over 24 hours (remember UK time is 1 hour ahead of UTC) till the world and his wife tries to break the WCG and Cosmology servers by unleashing bunkers.

Please ensure that the Pentathlon site is actually ready to start counting credits before you drop your load. We don't want any little accidents :p
 
jesus. having to set up more VM's. one server managed to get enough to last it a while were the other wont , I have to keep making VM's but with Linux its doing them a lot faster than the windows server. my overclocked I5 chews threw them quite quick so VM;s have been needed aswell
 
Back
Top Bottom