One sure fire way to run off users is to stop feeding them.

Now, now, let's be nice. Travis said he has put out a request on the admin forums for help in figuring it out.

---

Milkyway is a no good project! Or, should I say a no good credit project. It is a project. It has good credit. There is no good credit where there is no work. Maybe we should all discuss the credits and the limits some more. What's the point if you can't get any work?

Dude, take a chill pill and set up some backup projects! It looks like I'm not having any problems, on multiple CPUs even. See my next message.

---

Milkyway is a no good project! Or, should I say a no good credit project. It is a project. It has good credit. There is no good credit where there is no work. Maybe we should all discuss the credits and the limits some more. What's the point if you can't get any work?

Dude, take a chill pill and set up some backup projects! It looks like I'm not having any problems, on multiple CPUs even. See my next message.

Aye, I with you there Lloyd. It's best to remember that this is a test project. Things sometimes don't go smoothly, or else it wouldn't be in an 'alpha' stage. Travis is working on it and things get better, worse, better, best, hotter, colder, up, down, round and round, left foot in, left out, shake it all about...

I'm now running on BOINC 6.4.6, hopefully things might get better for my GPUs and the world seem a more delightful place for me...

---

Aye, I with you there Lloyd.

Thanks. It's nice to know that I'm not just a voice crying out in the wilderness. While I haven't had much occasion to interact with Admins, Travis seems to be exemplary in terms of technical capability and also being responsive to volunteers. He's also way too nice (and/or too busy) to spend much time reminding people of all the steps he's already announced that he's taking in order to try to fix the glitches.

I'm now running on BOINC 6.4.6, hopefully things might get better for my GPUs and the world seem a more delightful place for me...

Cool. I really should do some client upgrades, as I'm having some issues, though I put one machine online just this week, so it's probably running a fairly new client.

I saw on my team's optimized app board that some maniac managed a hack that allows every CPU core you have, and your GPUs to all process SETI at the same time. It looked to be a particular way he jiggered the client configuration, so it might work on other projects (maybe even non-CUDA) [evil grin].

Of course, you had better have some kind of PSU and cooling in your rig, and no banzai overclocks (CPU or GPU), if you are going to push it that hard. In fact, he warns about GPU overclocks, and I've seen similar warnings elsewhere concerning doing BOINC processing on your GPU.

---

All y'all are probably going to want to lynch me for saying this, and I have six computers hooked to this project, and there's no sign that any of them have had any recent problems downloading WUs from this project. I'm not posting this to be mean and evil but rather to give all of you some ideas, and Travis some information that might help figure out what the deal is here, and maybe Slicker can hang around and not go off to SETI, or Cosmology, or something.

Anyway, on to my observations:

• Although I have six computers attached, none is particularly powerful. In fact, one is painfully slow, and, there is another one that is even slower, to the extent I have to feed the hamster to keep it running. I wouldn't have bothered, except that I had to load Untangle on top of Ubuntu on it (instead of freestanding Untangle), and I figured: "what the heck - why not?"
  
• When I first attached to this project, I noticed that I sometimes had to suspend other projects to get it to load WUs (such as Renata/Neal mentioned). My thought was "...phooey on this!", so I jacked the resource share up to 10,000. Yes, you read that right. I've still had to do a manual update since then, but only once that I can recall.
  
• What counts for my most powerful rig (the Opty 170) has a number of projects attached, but most of them are on "No New Tasks" right now, notably PrimeGrid and SETI, which seem to download WUs without regard to resource share.
  
• I do have two backup projects set up on that machine: Leiden Classical and Spinhenge@home. I get quite a number of "no work sent...won't finish in time" warnings from Leiden (and little else), and it seems that Spinhenge isn't too proud to send me a WU here and there. Even though these have fairly low allocations (3 and 100, respectively), it seems to be adequate to keep me in WUs while I'm away or asleep, though (again) this isn't a particularly powerful CPU. And I don't have GPU that can process anything, much less this project. And, if we run into a situation where the system is just plain down for a time (like it was a while back), I can always allow it to download some Einstein or Rosetta.
  
• Note that, for my own reasons, I recently decided to load some Cosmology and yo-yo on one machine, and put Milkyway on NNT (it still had WUs, it was just working through them and not getting more). About half an hour ago, I allowed tasks on MW on that machine again, and it immediately downloaded a number of them, with no need to hit the update button. I checked just now, and it is trying to download work pretty frequently (automatically), but the server is complaining that I've reached my limit of 6 tasks per CPU. I wonder if that's because it's a dual Xeon box with HT, and the server has figured out that two of my CPUs are virtual.
  
• Oh, yeah. Four of them are linux boxes and I don't know that the benchmarks will run on any of them. I'm fairly certain that the benchmarks just error out on at least one of them (the one of the four that has an AMD processor).
  

---

In fact, he warns about GPU overclocks, and I've seen similar warnings elsewhere concerning doing BOINC processing on your GPU.

I'm really glad you mentioned that. I've found my PC with GPU turned off 3 times over this w/e. I suspect it's cut off becuase of the GPU leaving the last MW WU with a computation error. Otherwise no other lost or bad WU's that I can see. I've now taken the GPU back to the default settings, also because ever-so-slightly faster than extrememly fast is not going to make that much of a difference I suspect.

I've now installed BOINC 6.4.6 on a Vista machine, which seems to be optimizing nicely.

---

• When I first attached to this project, I noticed that I sometimes had to suspend other projects to get it to load WUs (such as Renata/Neal mentioned). My thought was "...phooey on this!", so I jacked the resource share up to 10,000. Yes, you read that right. I've still had to do a manual update since then, but only once that I can recall.
  

I've had MilkyWay set to 10,000 resource share (99.01%) with CPDN to 100 (0.99%), and also MilkyWay 10000, GPUGRID 100, CPDN 100 - for the past 2 weeks plus.

I think you have been lucky, evil and mean if you havn't been pounding that 'update' button like the rest of us ;)

---

I think you have been lucky, evil and mean if you havn't been pounding that 'update' button like the rest of us ;)

Yeah, right. I'm glad to see that this hasn't affected your sense of humor, at least.

All kidding aside, there has to be a rational explanation here.

Oh, yeah - I have found that, recently, on a couple of my boxes, there have been completed WUs that hadn't been uploaded. Come to think of it, I've used "Update" to take care of that. But I've always gotten replacement WUs right away, with no problem.

The only other thing I could think of was BOINC client version, which has to be all over the map on my farm.

---

I have found that boinc 6.5.0 seems to be the most stable platform to run GPU projects from......I have several nvidia cards and one ATI card and they are all running along smoothly..........

Been receiving a nice constant stream of work for 2 days now, without and idle boxes or UPDATE button stabbing to worry about!! This, IMO, is due to the increase in WU length.....knock-on effect: gives the scheduler queue time to reach FULL, instead of emptying and then refilling hiccup style from the feeder as Travis previously mentioned ( not in those exact words, but the meaning's the same ). Also keeping the 6xCPU WU limit in place seems to have contributed to containing the constant drain from the GPU clients...

All in all, if the server can carry on with hundreds of thousands of HTTP requests per day, then the prob. seems to be fixed.

BTW, Travis, how are the results from the new stripes (24, 25, and 26) looking?

---

Seejay **Proud Member and Founder of BOINC Team Allprojectstats.com**

Oh, yeah - and one of them doesn't have an update button, because I'm running Ubuntu server, which is strictly CLI. Of course there is a way to force updates, and it's frankly kind of a pain.

---

Oh, yeah - and one of them doesn't have an update button, because I'm running Ubuntu server, which is strictly CLI. Of course there is a way to force updates, and it's frankly kind of a pain.

What!!??

You've got other hosts running, so why should your CLI Linux host be any trouble or real effort to manage?

You can access it with BOINC Manager from any machine as long as you use the default RPC port. If you've changed it, you can use BOINCView from a Winbox.

Alinator

Alinator wrote:

Lloyd wrote:

Oh, yeah - and one of them doesn't have an update button, because I'm running Ubuntu server, which is strictly CLI. Of course there is a way to force updates, and it's frankly kind of a pain.

What!!??

You've got other hosts running, so why should your CLI Linux host be any trouble or real effort to manage?

You can access it with BOINC Manager from any machine as long as you use the default RPC port. If you've changed it, you can use BOINCView from a Winbox.

Alinator

What he said!! ;^).

---

Seejay **Proud Member and Founder of BOINC Team Allprojectstats.com**

You've got other hosts running, so why should your CLI Linux host be any trouble or real effort to manage?

I have Boincview. It's the CLI that's a pain.

---

All y'all can stop hating me now:

3/2/2009 5:43:26 PM|Milkyway@home|Computation for task ps_s23_11_586182_1236020433_0 finished
3/2/2009 5:43:26 PM|Milkyway@home|Starting ps_s23_11_586184_1236020433_0
3/2/2009 5:43:26 PM|Milkyway@home|Starting task ps_s23_11_586184_1236020433_0 using milkyway version 19
3/2/2009 5:43:28 PM|Milkyway@home|Started upload of ps_s23_11_586182_1236020433_0_0
3/2/2009 5:43:30 PM|Milkyway@home|Finished upload of ps_s23_11_586182_1236020433_0_0
3/2/2009 5:43:34 PM|Milkyway@home|Sending scheduler request: To report completed tasks.  Requesting 0 seconds of work, reporting 1 completed tasks
3/2/2009 5:43:39 PM|Milkyway@home|Scheduler request succeeded: got 0 new tasks
3/2/2009 5:44:26 PM|Milkyway@home|Computation for task ps_s23_11_586181_1236020433_0 finished
3/2/2009 5:44:26 PM|Milkyway@home|Starting ps_s23_11_586185_1236020433_0
3/2/2009 5:44:26 PM|Milkyway@home|Starting task ps_s23_11_586185_1236020433_0 using milkyway version 19
3/2/2009 5:44:28 PM|Milkyway@home|Started upload of ps_s23_11_586181_1236020433_0_0
3/2/2009 5:44:31 PM|Milkyway@home|Finished upload of ps_s23_11_586181_1236020433_0_0
3/2/2009 5:44:34 PM|Milkyway@home|Sending scheduler request: To report completed tasks.  Requesting 0 seconds of work, reporting 1 completed tasks
3/2/2009 5:44:39 PM|Milkyway@home|Scheduler request succeeded: got 0 new tasks
3/2/2009 5:46:24 PM|Leiden Classical|Sending scheduler request: To fetch work.  Requesting 513 seconds of work, reporting 0 completed tasks
3/2/2009 5:46:29 PM|Leiden Classical|Scheduler request succeeded: got 0 new tasks
3/2/2009 5:46:29 PM|Leiden Classical|Message from server: No work sent
3/2/2009 5:46:29 PM|Leiden Classical|Message from server: (won't finish in time) Computer on 99.3% of time, BOINC on 100.0% of that, this project gets 0.0% of that
3/2/2009 5:49:34 PM|Leiden Classical|Sending scheduler request: To fetch work.  Requesting 513 seconds of work, reporting 0 completed tasks
3/2/2009 5:49:39 PM|Leiden Classical|Scheduler request succeeded: got 0 new tasks
3/2/2009 5:49:39 PM|Leiden Classical|Message from server: No work sent
3/2/2009 5:49:39 PM|Leiden Classical|Message from server: (won't finish in time) Computer on 99.3% of time, BOINC on 100.0% of that, this project gets 0.0% of that
3/2/2009 5:52:44 PM|Leiden Classical|Sending scheduler request: To fetch work.  Requesting 514 seconds of work, reporting 0 completed tasks
3/2/2009 5:52:49 PM|Leiden Classical|Scheduler request succeeded: got 0 new tasks
3/2/2009 5:52:49 PM|Leiden Classical|Message from server: No work sent
3/2/2009 5:52:49 PM|Leiden Classical|Message from server: (won't finish in time) Computer on 99.3% of time, BOINC on 100.0% of that, this project gets 0.0% of that
3/2/2009 5:55:40 PM|Milkyway@home|Sending scheduler request: Requested by user.  Requesting 0 seconds of work, reporting 0 completed tasks
3/2/2009 5:55:45 PM|Milkyway@home|Scheduler request succeeded: got 0 new tasks
3/2/2009 5:55:55 PM|Leiden Classical|Sending scheduler request: To fetch work.  Requesting 514 seconds of work, reporting 0 completed tasks
3/2/2009 5:56:00 PM|Leiden Classical|Scheduler request succeeded: got 0 new tasks
3/2/2009 5:56:00 PM|Leiden Classical|Message from server: No work sent
3/2/2009 5:56:00 PM|Leiden Classical|Message from server: (won't finish in time) Computer on 99.3% of time, BOINC on 100.0% of that, this project gets 0.0% of that
3/2/2009 5:56:40 PM|Milkyway@home|Sending scheduler request: Requested by user.  Requesting 0 seconds of work, reporting 0 completed tasks
3/2/2009 5:56:45 PM|Milkyway@home|Scheduler request succeeded: got 0 new tasks
3/2/2009 6:02:05 PM|Milkyway@home|Sending scheduler request: Requested by user.  Requesting 0 seconds of work, reporting 0 completed tasks
3/2/2009 6:02:10 PM|Milkyway@home|Scheduler request succeeded: got 0 new tasks

...etc. I'll see what the other boxes are doing...

---

Meanwhile, it finally decided to ask for some work, and got some, all by itself.

3/2/2009 6:42:31 PM|Milkyway@home|Computation for task ps_s23_11_586184_1236020433_0 finished
3/2/2009 6:42:31 PM|Spinhenge@home|Starting 10_Fe30_map_1050_52_1
3/2/2009 6:42:31 PM|Spinhenge@home|Starting task 10_Fe30_map_1050_52_1 using metropolis version 312
3/2/2009 6:42:31 PM|Milkyway@home|Sending scheduler request: To fetch work.  Requesting 1532520 seconds of work, reporting 0 completed tasks
3/2/2009 6:42:33 PM|Milkyway@home|Started upload of ps_s23_11_586184_1236020433_0_0
3/2/2009 6:42:36 PM|Milkyway@home|Finished upload of ps_s23_11_586184_1236020433_0_0
3/2/2009 6:42:36 PM|Milkyway@home|Scheduler request succeeded: got 10 new tasks
3/2/2009 6:42:38 PM|Milkyway@home|Started download of 86-v5.txt
3/2/2009 6:42:38 PM|Milkyway@home|Started download of ps_s86_11_search_parameters_755458_1236044431
3/2/2009 6:42:39 PM|Milkyway@home|Finished download of 86-v5.txt
3/2/2009 6:42:39 PM|Milkyway@home|Finished download of ps_s86_11_search_parameters_755458_1236044431
3/2/2009 6:42:39 PM|Milkyway@home|Started download of ps_s82_11_search_parameters_755423_1236044431
3/2/2009 6:42:39 PM|Milkyway@home|Started download of ps_s82_11_search_parameters_755424_1236044431
3/2/2009 6:42:40 PM|Milkyway@home|Finished download of ps_s82_11_search_parameters_755423_1236044431
3/2/2009 6:42:40 PM|Milkyway@home|Finished download of ps_s82_11_search_parameters_755424_1236044431
3/2/2009 6:42:40 PM|Milkyway@home|Started download of ps_s82_11_search_parameters_755425_1236044431
3/2/2009 6:42:40 PM|Milkyway@home|Started download of ps_s82_11_search_parameters_755426_1236044431
3/2/2009 6:42:40 PM|Milkyway@home|Starting ps_s86_11_755458_1236044431_0
3/2/2009 6:42:40 PM|Milkyway@home|Starting task ps_s86_11_755458_1236044431_0 using milkyway version 19
3/2/2009 6:42:41 PM|Milkyway@home|Finished download of ps_s82_11_search_parameters_755425_1236044431
3/2/2009 6:42:41 PM|Milkyway@home|Finished download of ps_s82_11_search_parameters_755426_1236044431
3/2/2009 6:42:41 PM|Milkyway@home|Started download of ps_s82_11_search_parameters_755427_1236044431
3/2/2009 6:42:41 PM|Milkyway@home|Started download of ps_s82_11_search_parameters_755429_1236044431
3/2/2009 6:42:42 PM|Milkyway@home|Finished download of ps_s82_11_search_parameters_755427_1236044431
3/2/2009 6:42:42 PM|Milkyway@home|Finished download of ps_s82_11_search_parameters_755429_1236044431
3/2/2009 6:42:42 PM|Milkyway@home|Started download of ps_s82_11_search_parameters_755430_1236044431
3/2/2009 6:42:42 PM|Milkyway@home|Started download of ps_s82_11_search_parameters_755431_1236044431
3/2/2009 6:42:43 PM|Milkyway@home|Finished download of ps_s82_11_search_parameters_755430_1236044431
3/2/2009 6:42:43 PM|Milkyway@home|Finished download of ps_s82_11_search_parameters_755431_1236044431
3/2/2009 6:42:43 PM|Milkyway@home|Started download of ps_s82_11_search_parameters_755432_1236044431
3/2/2009 6:42:44 PM|Milkyway@home|Finished download of ps_s82_11_search_parameters_755432_1236044431
3/2/2009 6:42:47 PM|Milkyway@home|Sending scheduler request: To report completed tasks.  Requesting 0 seconds of work, reporting 1 completed tasks
3/2/2009 6:42:52 PM|Milkyway@home|Scheduler request succeeded: got 0 new tasks
3/2/2009 6:43:44 PM|Milkyway@home|Computation for task ps_s23_11_586185_1236020433_0 finished
3/2/2009 6:43:44 PM|Milkyway@home|Starting ps_s82_11_755423_1236044431_0
3/2/2009 6:43:44 PM|Milkyway@home|Starting task ps_s82_11_755423_1236044431_0 using milkyway version 19
3/2/2009 6:43:46 PM|Milkyway@home|Started upload of ps_s23_11_586185_1236020433_0_0
3/2/2009 6:43:48 PM|Milkyway@home|Finished upload of ps_s23_11_586185_1236020433_0_0
3/2/2009 6:43:52 PM|Milkyway@home|Sending scheduler request: To report completed tasks.  Requesting 0 seconds of work, reporting 1 completed tasks
3/2/2009 6:43:57 PM|Milkyway@home|Scheduler request succeeded: got 0 new tasks

Yes, I am confused. I've found some reports of goofiness with the BOINC 6.4.X client, which, unfortunately, my latest one is.

My Opty is still running (would you believe) 5.10.28

---

Alinator wrote:

You can access it with BOINC Manager from any machine as long as you use the default RPC port. If you've changed it, you can use BOINCView from a Winbox.

Seejay wrote:

What he said!! ;^).

Maybe not so much. Given the evidence from troubleshooting is that the newer client can select a computer running an older client, but not the other way around - e.g. in BOINC manager, my Netvista (6.4.6) can select my Opty (5.10.28), but not the other way around, even with the firewall turned off in the former.

And it appears that Boincview has been abandoned/orphaned entirely. The web site is down and nobody seems to know what happened to its author. What I have found is reports of it not playing nice with newer BOINC clients, and that seems to be the case based on personal experience. I wonder if it's a matter of Boincview looking for data where it no longer is, as they moved some stuff with 6.X

---

you could use the boinccmd command to issue an update with
./boinccmd --project http://milkyway.cs.rpi.edu/milkyway/ update

alias that to something easier, say upd_milky with
alias upd_milky='~/BOINC/boinccmd --project http://milkyway.cs.rpi.edu/milkyway/ update'
you'll just need maybe to substitue ~/BOINC with the path to where boinccmd lives on your system

Temujin wrote:

you could use the boinccmd command to issue an update with
./boinccmd --project http://milkyway.cs.rpi.edu/milkyway/ update

alias that to something easier, say upd_milky with
alias upd_milky='~/BOINC/boinccmd --project http://milkyway.cs.rpi.edu/milkyway/ update'
you'll just need maybe to substitue ~/BOINC with the path to where boinccmd lives on your system

Cool. Thank You. That's very constructive and useful advice. I had been puzzling out the update command, and then stepping back through the command queue to reuse it. I wasn't aware of the alias capability

Back in the day, with DOS, I used to create batch files directly from the console, because you couldn't count on anything other than the truly horrible edlin to be present. Later, I started carrying several utilities around on a floppy disk (imagine THAT), one of which would be Qedit, which was full screen and responded to WordStar commands.

I really should learn some shell programming - I imagine I could handle it, even at my ripe old age ;^)

Thanks again.

---

./boinc_cmd --help

{edit}

also BOINC command tool help page

{/edit}