Message boards :
Number crunching :
3 Computers Running Seti v7 Not Automaticaly Reporting Work
Message board moderation
Author | Message |
---|---|
Lee Gresham Send message Joined: 12 Aug 03 Posts: 159 Credit: 130,116,228 RAC: 0 |
This is a minor problem I've not seen since the server move until yesterday. To alleviate the problem I started using Task Scheduler last night to force work reporting once an hour. The 3 pcs are Windows XP x64 boxes and all are wireless to DSL modem. A Win7 box connected thru LAN cable is not affected. After updating to newest Boinc version and setting up for Seti v7 all was fine for 5 about days. I noticed early yesterday each remote pc had 20 or more completed work units waiting to be reported. This was common before the Seti server move and I used Task Scheduler then to manage reporting. Checking the event log, all reporting by the 3 has been "Requested By User". No auto updating has occurred. Delta-V |
Fred E. Send message Joined: 22 Jul 99 Posts: 768 Credit: 24,140,697 RAC: 0 |
This is a minor problem I've not seen since the server move until yesterday. To alleviate the problem I started using Task Scheduler last night to force work reporting once an hour. The 3 pcs are Windows XP x64 boxes and all are wireless to DSL modem. A Win7 box connected thru LAN cable is not affected. After updating to newest Boinc version and setting up for Seti v7 all was fine for 5 about days. I noticed early yesterday each remote pc had 20 or more completed work units waiting to be reported. This was common before the Seti server move and I used Task Scheduler then to manage reporting. If you still have bad time estimates, you may need to increase the cache settings so BOINC doesn't think you have too much work. Also make sure you don't have any suspended tasks or stuck file transfers. The option <report_results_immediately>1</report_results_immediately> in the options section of cc_config.xml will force that reporting. Another Fred Support SETI@home when you search the Web with GoodSearch or shop online with GoodShop. |
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4 |
This is a minor problem I've not seen since the server move until yesterday. To alleviate the problem I started using Task Scheduler last night to force work reporting once an hour. The 3 pcs are Windows XP x64 boxes and all are wireless to DSL modem. A Win7 box connected thru LAN cable is not affected. After updating to newest Boinc version and setting up for Seti v7 all was fine for 5 about days. I noticed early yesterday each remote pc had 20 or more completed work units waiting to be reported. This was common before the Seti server move and I used Task Scheduler then to manage reporting. Lee, What are your cache settings, both of them? The wording are on the Setiathome Computing Preferences page: Maintain enough tasks to keep busy for at least --- days Or in the 7.0.xx Boinc Manager Preferences page: Minimum work buffer --- days Claggy |
Keith White Send message Joined: 29 May 99 Posts: 392 Credit: 13,035,233 RAC: 22 |
I went from a 6 to 10 day queue and that really didn't do a whole lot. Currently I have 0 MB V6 (as expected) and 92 MB V7 split 42 CPU and 50 GPU. The CPU is now using server estimates while the GPU is using a <flops> guesstimate since once again my GPU validated, non-overflow results aren't being counted as "completed" or has a processing rate computed, just like what happen with MB V6 (over 10,000 WU done, only 9 counted as completed, what bollocks). I am however now entertaining AP workunits for the first time. It will be a while until enough of either CPU (4 done, 2 validated, 1 "completed" and 3 in progress with the current guesstimate is off by 10x) or GPU (none attempted yet, 2 in progress) are done to get proper server side guesstimates. So I don't know if those are counting against my 100/100. Not that I'm desperately needing more units. As it is MV V7 CPU has roughly 2 1/3 days of work while the GPU has 3 1/8 days. The AP adds about 0.9 days to the CPU total and no data for GPU AP yet. Now in the past with V6 when I got 100/100 WUs it turned out to be 3 to 4 days worth of work. So nothing really changed from a queue depth in days (I was asking for 6 then) except now even bumping it up to 10 I'm still only getting less than 4 days worth of work. Just another oddity with the whole V7 switch over. Edit: Another oddity is reporting finished WUs. Sometimes it's immediate, sometimes it's more than 6 hours, building up a pile to be reported all at once. Queue set to 10/-. "Life is just nature's way of keeping meat fresh." - The Doctor |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
I went from a 6 to 10 day queue and that really didn't do a whole lot. Claggy was asking exactly which setting you changed - website or BOINC Manager, and which of the two values available in each place. There are subtle interactions between the two/four that you may not be aware of - best to give the whole picture. |
Keith White Send message Joined: 29 May 99 Posts: 392 Credit: 13,035,233 RAC: 22 |
I went from a 6 to 10 day queue and that really didn't do a whole lot. Web site obviously since it's location in the BOINC Manager is a secret known only to it's hooded initiates. ;) OK now they match (well 10/0 in the BOINC Manager, was 6/0.1), a request was made, WUs reported and nothing came of it since we've run dry of WUs to distribute. So To Be Continued once the Spice flows again. Edit: Minor rant here. What the heck is the purpose then of the Your preferences button in the BOINC manager that brings up the web preferences that allows you to set your queue depth, which then tells you to do a project update to transfer that info if it is then overridden by a local setting in the BOINC manager that doesn't get updated with the update? Why? And why is it the first 2 items on the web network preferences but the third line on the network tab in the BOINC manager? Eyes glazed over looking at transfer rates and time blocks and overlooked "Minimum Work Buffer". What, "days of work" too straightforward a description? Let me guess, the UI design was done by a former SAP forms developer. :eyeroll: "Life is just nature's way of keeping meat fresh." - The Doctor |
Lee Gresham Send message Joined: 12 Aug 03 Posts: 159 Credit: 130,116,228 RAC: 0 |
Glad I checked the cache preferences! Web account page was fine. 10 Days minimum & 10 Days maximum. The Boinc Manager computing preferences were another matter. Of the 3 pcs all had the maximum set to 10 days but of the minimums one was set to 0.01 days and the other two were set to 0 days. I noticed some of the other settings set differently from my usual preferences t0o. A fluke in the 7.0.64 installer? All 3 were upgraded from the Boinc 6.1x series early this week. I'll have to watch preferences more carefully after new upgrades. As soon as I changed the settings to 10/10 the computers began downloading work too. Two were nearly out of work and the other only had about half a normal cache. Thanks for all the help.................. Delta-V |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13727 Credit: 208,696,464 RAC: 304 |
All 3 were upgraded from the Boinc 6.1x series early this week. There's the problem. The settings between BOINC v6 & v7 cache settings were reversed. When going from v6 to v7 you have to swap the values around to maintain the same level of cache. Keep in mind there are some issues with the splitters at the momement so there's not a lot of work available at present, although the amount available has been increasing over the last couple of hours, but at a glacial pace. EDIT- keep in mind setting such large cache settings will also limit the amount of work the systems will download, especially with the much longer v7 run times. Even then, the maximum amount of work you can get is 100WUs for the CPU & 100 WUs for the GPU- these are server side limits. Grant Darwin NT |
Lee Gresham Send message Joined: 12 Aug 03 Posts: 159 Credit: 130,116,228 RAC: 0 |
There are subtle interactions between the two/four that you may not be aware of - best to give the whole picture.[/quote] Web site obviously since it's location in the BOINC Manager is a secret known only to it's hooded initiates. ;) In the Boinc Manager click on the Tools tab at the page top and select Computing Preferences from the drop down. Delta-V |
Gatekeeper Send message Joined: 14 Jul 04 Posts: 887 Credit: 176,479,616 RAC: 0 |
When I made the switch from BOINC 6.12.34 to 7.0.64, I was getting only about 75% of the "limit" across my three rigs, even though I had the "right" settings on each rig's local preferences. I finally checked general preferences on the web, and they were totally sideways. Correcting them solved my problem. So, I gather that general preferences now override local preferences, instead of vice versa. |
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4 |
Glad I checked the cache preferences! Web account page was fine. 10 Days minimum & 10 Days maximum. The Boinc Manager computing preferences were another matter. Of the 3 pcs all had the maximum set to 10 days but of the minimums one was set to 0.01 days and the other two were set to 0 days. I noticed some of the other settings set differently from my usual preferences t0o. A fluke in the 7.0.64 installer? All 3 were upgraded from the Boinc 6.1x series early this week. I'll have to watch preferences more carefully after new upgrades. As soon as I changed the settings to 10/10 the computers began downloading work too. Two were nearly out of work and the other only had about half a normal cache. To add to what Grant has said, Boinc 7 has a totally new scheduler, where with Boinc 6 it would fill to the Min and Max amoints, and keep at that level, Boinc 7 will fill to the Min and Max amounts, then wait for the amount of work in the cache to drop befow the Min cache setting before asking again, Having too high a cache setting doesn't help either, with 10 + 10 you're trying to get 20 days work, but the first Cache setting also says you're going to be unavailable for 10 days, so that work needs to be done 10 days earlier, with Shortie deadlines being 21 days now, you're trying to get 20 days work done in 11 days, the server either won't send you work, or you won't get it all done in time, especially if we get a Shortie storm and Boinc 7 keeps filling up with that type of work. Best to go for a Meduim level cache setting of something like 5 + 0.01 for Boinc 7 (you have 4 venues and local preferences available to split Boinc 6 hosts off to), Boinc will fill to 5.01 days work, then every time it drops below 5 days it'll ask for work. Claggy |
Keith White Send message Joined: 29 May 99 Posts: 392 Credit: 13,035,233 RAC: 22 |
So this morning I got up and I had 14 results waiting to be reported. 9 1/2 hours since the previous schedule request which I did before going to bed. I thought that the manager had at least a maximum time between requests if results are available to report or does it only do that if it decides to need more work? Could this have been caused as a side effect of the drought when I did get AP WUs and since the server doesn't yet have a good estimate of my AP performance so it thinks I have loads of work to do? I wouldn't think that has anything to do with not reporting, unless it now only reports when it thinks you are running low of work. I'm getting flashbacks to the days where I manually connected on dial-up to report and hope that the server supply hadn't run dry. "Life is just nature's way of keeping meat fresh." - The Doctor |
John Clark Send message Joined: 29 Sep 99 Posts: 16515 Credit: 4,418,829 RAC: 0 |
You could apply a batch file, in the Command window, which updates the project every 2 completed WUs. The example tries to update SETI WUs every 1350 seconds (22.5 minutes. This time, in seconds, can be altered in NotePad and saved. To use it as a .Bat file that extension needs to be used. It will in the BOINC folder to activate the Command window. For x64 OS the folder is different. An example might be - cls :START echo off time /t PING -n -w 1000 1350 127.0.0.1>nul boinccmd --project http://setiathome.berkeley.edu/ update GOTO START It's good to be back amongst friends and colleagues |
Keith White Send message Joined: 29 May 99 Posts: 392 Credit: 13,035,233 RAC: 22 |
Well adding a <flops> line into the CPU AP app_info block to reflect the value of the one AP unit "completed" (a temporary measure until 11 are) unjammed the CPU WUs. I'm now at 5 AP CPU and 95 MB V7 CPU. GPU is still jammed at 5 AP and 47 MB V7 but I have zero info on the effectiveness of the GPU AP app relative to the CPU one so I'm not touching anything on that side of things until I have some actual data to base a guess on. Now that I've hit the 100 CPU task limit, it's asking for more WU every 5 minutes now so slow reporting shouldn't be a problem anymore. "Life is just nature's way of keeping meat fresh." - The Doctor |
Gundolf Jahn Send message Joined: 19 Sep 00 Posts: 3184 Credit: 446,358 RAC: 0 |
I thought that the manager had at least a maximum time between requests if results are available to report or does it only do that if it decides to need more work? Yes, that's 24 hours. So, you've only waited a third of the necessary time. ;-) Gruß, Gundolf |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
The example tries to update SETI WUs every 1350 seconds (22.5 minutes. No, unfortunately it doesn't. The -n and -w switches each require a numeric argument. A better rendering would be PING -n 1350 -w 1000 127.0.0.1>nul |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.