8th Annual BOINC Pentathlon

[StefanR5R]

Some time before my initial bunker is depleted, I will set up 6 client instances at once, each for 1 GPU feeder. Two configured to use GPU 0 only, two for GPU 1, two for GPU 2. My hope is that each of these clients will receive enough tasks that I can take hands off until May 9.

Besides, I want to share the host's 10-core CPU for Cosmology. Not sure yet whether I'll set the GPU feeders' process priority lower or higher than the Cosmology VM tasks...

 

[TennesseeTony]

1h40m until the race starts, or is SUPPOSED to start. If you are planning to release your bunkers and get fresh tasks, I'd wait until I see some points on the board first. There have been issues in the past, I'd hate for anyone to have so much work not get counted for the race.

 

[crashtech]

Really need to get my machines back to work, hate just letting them sit there full of WUs. This was a pretty major miscalculation.

 

[GLeeM]

Is everyone sitting there clicking update waiting for someone else to post? click, click, click!!

When will "I see some points on the board first"? After an hour? Trying to have patience

 

[Ken g6]

Is everyone sitting there clicking update waiting for someone else to post? click, click, click!!

When will "I see some points on the board first"? After an hour? Trying to have patience

They posted in the shoutbox that we can upload. Apparently they're using the WCG race system. And we can upload cosmology too.

 

[crashtech]

I'm going to have to flush one soon, I just sold the guts out of it and they're coming over to get it lol.

 

[TennesseeTony]

Well the chat bar on the side of the stats page has a message or three from SG, saying they are go for recording the stats, Cosmo first (every hour), but WCG is not as frequent. Your systems can go pee now if they can't hold it any longer.

EDIT: Ken beat me to it.

 

[TennesseeTony]

Looks like Cosmo server is buggered. WCG uploading just fine.

 

[iwajabitw]

Sorry for not having any horses in the CPU race, but I'm bunkering Einstein up just fine.

 

[Ken g6]

WCG reporting they're out of zika WUs for Linux.

 

[TennesseeTony]

And Windows.

 

[Ken g6]

Huh, got some cosmology work, but how do I get it to run only one core per WU? I have plenty of RAM.

 

[Orange Kid]

Some stats are up

I'm getting Zika work just fine here. Cosmology I still have as I've reduced my Q size there

 

[Markfw]

well, I aborted a lot of tasks, but I can;t get only Open zika, but trying.....

 

[TennesseeTony]

For those who love stats, and don't yet know, you can click any team name to get what I think are the best stats of any race. Currently we are a whopping 20 points ahead of our friends over at OcUK in Zika.

Edit: I don't know if my Cosmo tasks will be done uploading before this race ends, lol. Taking their sweet time exiting my computers. But the good news is that, well, we have (substantially?) more points on the way to that project.

 

[Markfw]

So, when I get time, I need to change back to 1/2 day or so instead of 3. This bunkering killed my Open Zika contribution from the last race.

You guys need to clue me in to all this stuff way in advance. Another race that buggered for us.

 

[Ken g6]

So, when I get time, I need to change back to 1/2 day or so instead of 3. This bunkering killed my Open Zika contribution from the last race.

You guys need to clue me in to all this stuff way in advance. Another race that buggered for us.

Mark, I think you might be taking this racing too seriously, and being too hard on yourself. It seems like I almost always perform less well than I could have, or bunker less than I should. I take it as a learning experience, but I don't worry too much about it. It really seems to take time and practice for me to learn this stuff. You've had less of both than I have with Boinc, and like I said I'm not perfect either. but we'll get there.

Regarding the specific issue of getting other than zika work, did you see Tony's picture on the last page? You might also click the device manager link to make sure none of your device profiles have different settings.

 

[StefanR5R]

About blocking uploads with C:\Windows\System32\divers\etc\hosts:

The host file trick will be the way to go I think

Ya, it's windows, ya , you gotta reboot.

I didn't need to reboot on Win 7 Pro.
(I know, nobody around here has Win 7...)

For the record, in order to unblocked the servers again, I simply prepended the respective lines in the hosts file with a # sign. Then I tested to get to the projects' sites with the web browser, and it worked. And boinc began uploading just fine as well.

I.e. no reboot needed on Win 7 at least.

Edit:
The client PC itself is configured to a fixed own IP address and fixed DNS server addresses. I.e. no DHCP. I don't know whether or not this matters to how Windows reacts to on-the-fly changes of the hosts file.

 

[StefanR5R]

Current Einstein team stats indicate some leaks in our bunkers.
http://stats.free-dc.org/stats.php?page=team&proj=eah&team=12&sort=yesterday

 

[crashtech]

Yeah, that would be me. If there is a way to selectively bunker projects, I haven't learned it yet

 

[iwajabitw]

as of this morning Einstein for me was still not taking any uploads or giving me any new tasks. The 980's have about 900 bunkered that should be good up to a few hours before the race starts. My 280x's though only got 704 tasks and they complete 12 per hour so I only have about 3 days worth. Event log was telling me there was no work available and on there status page a few servers were not running. Are any of you able to get any new tasks today?

Edit: Uploads went, but no new tasks.Event log still saying No jobs available.

 

[Orange Kid]

I had a leak too. Must have forgot to reboot a computer. Yes Win 10 needs to reboot to read the host file

 

[StefanR5R]

Yeah, that would be me. If there is a way to selectively bunker projects, I haven't learned it yet

In order of amount of work to do, from least to most:

1. Define nonsensical IP addresses for the project servers in the hosts file.
2. Block outbound traffic to the project servers in a "personal firewall" or external firewall.
3. Launch separate client instances per project.
4. (edit) Run a background script which selectively suspends tasks.

The hosts to block in 1. or 2. are those listed in client_state.xml in the boinc data directory (default C:\ProgramData\BOINC).

Pitfalls:

• Method 1. has the downside that it also blocks the web browser from accessing the project site.
  (edit) As @Orange Kid notes, a big downside on Win 10 is that reboots may be required to apply changes to the hosts file.
  (edit) @TennesseeTony reports that a few minutes waiting for the change to become active on Win 10 was sufficient for him.
  And @Ken g6 pointed out that some Linux distros are set up to cache hosts-file based lookups too.
  
• Method 2. has the same downside, unless you use a personal firewall in which you can specify programs which are to be blocked. It furthermore has the downside that you first have to look up the IP addresses of the servers, which may even be ranges, and could change over time. (Use "nslookup" to translate from server name to IP address.)
• Method 3., relying on disabling network access in the desired boinc client instance, does not work for vboxwrapper applications like Cosmology and LHC.

Meanwhile I explored method 1. successfully on Windows, and method 3. on Linux and Windows.

Edit, May 6: Added info from Ken g6 and TennesseeTony.

 

[Ken g6]

@StefanR5R, then there's method 4, my automated suspension method. You only want to do that on projects with predictable runtimes - though I could probably find another way. It would take a rewrite to work in Windows. And you either want a GPU project or you want to turn off leaving tasks in memory while suspended.

 

[StefanR5R]

@Ken g6, indeed, how could I forget this suspend trick.
(Doesn't actually need a rewrite for Windows though; just requires bash and friends.)