Leveraging, did you think is possible, in the near future, to add the total numbers of WU with pending credits as i explain before? Thats one thing i realy miss from SETI Boinc, a way to know easely the total production of the entire fleet/host, that number gives us a fast way to know if all ships are working fine without the need to go ship by ship asking if everything is OK. Just the number of WU, don´t whant to overload any server database.
I really don't have a satisfactory answer to your question. I understand what you want and, yes, it would be nice to have those numbers visible for the various categories of tasks when you look at the tasks list for a particular host. I also saw your suggestion for a 'pendings summary' giving the total number of pendings for a particular host. The Devs will have seen it and it could be something they would have a look at at some stage.
The problem is that the server code here is highly customised for the needs of this particular project. I think that E@H is the only project using 'Locality Scheduling' and many of the customisations to support that were developed and maintained here and this would all have to be retro-fitted each time if the project tried to keep up with the latest standard BOINC server code. As is always the case, the Dev team is small and overworked so even minor enhancements can take a while to be implemented when there are so many other fires to fight.
Personally, I use a different strategy to make sure all hosts are operating normally and for spotting any that have failed in any way. I don't even try to watch numbers of pendings, valids, invalids or errors directly. I use RAC and time of last contact. I have a lot of hosts and I know roughly what the RAC of each one should be. I know all about why RAC is a bad measure of performance (and I largely agree) but I've actually found it to be rather useful in spotting changes in performance despite this.
I watch the complete list of all my hosts which is arranged in order from 'most recent contact' to 'oldest contact'. If everything is OK, the bottom host on the list will be at a 'believable' value for last contact. If any seem to be a bit too long ago, I look at just those hosts and fix whatever the problem is (if any). Sometimes the machine may have crashed so just a reboot but quite often it's still crunching with the network connection down. Also sometimes a particular task just takes longer than usual so this blows the DCF out to the point that the host thinks it doesn't need more work for quite a while. I keep a 3.5 day cache on many hosts so a task taking 8 hours instead of 5 will cause the host to think it now has a 5.6 day cache and it wont initiate a contact for quite a while until the DCF is slowly worked down again.
I try to check the hosts list a couple of times per day but sometimes I can't and so the system isn't perfect. I'm not always physically there where the hosts are so that makes it difficult to poke and prod when necessary. The machines are largely not in air-conditioned surroundings so now that it's summer here, heat stress failures are likely to be a lot more prevalent. I've already started cleaning fans/heatsinks and replacing thermal paste where necessary.
I try to check the hosts list a couple of times per day but sometimes I can't and so the system isn't perfect. I'm not always physically there where the hosts are so that makes it difficult to poke and prod when necessary. The machines are largely not in air-conditioned surroundings so now that it's summer here, heat stress failures are likely to be a lot more prevalent. I've already started cleaning fans/heatsinks and replacing thermal paste where necessary.
Thanks, and welcome to my world, to many ships to take care, imagine i only crunch in 8 hosts to try not to explode the power bill, actualy i have more 20 light ships - with no GPUs - ready to dutty all time.
No AC of course and a very hot enviroment, Brazil = Tropical Hot/wet country, begining of Summer here to. Thanks my city is located 1000 mts high so not extrealy hot (rises to +/- 36C on hot days). But with to many hosts arroud, heat is "the problem" and some days we literaly "fry on hell" and need to shoot/slow down most of the GPU´s for a while. Or when rain to much, heavy lighting (my place is literaly over a big iron mine), so even with massive PSU - lighting is allways have the potential to make the hosts go to some "frozen wierd state"
And a very few time to babysitting. Thats way i need to find an easy way to check all the hosts (crunches or no) easely. Until them, will stay making the old hi-low game to find the number.
Bye, will not take more of your precious time, see you around
I also use the 'last contact' method to try to keep tabs on my 9 rigs.
If one of them has not reported within a reasonable amount of time, I then have a look-see to make sure nothing has failed.
I try to find in the Boinc Manual but i was unable to do that.
Does anybody knows how to do this?
E@H does not need a proxy, SETI needs a proxy, How i configure that?
Of course i know how to make the boinc use a proxy, i just don´t know how to make it use a proxy for the SETI project and don´t use the proxy for the E@H project.
I see something in the configuration option of the proxy: don´t use for ??? just don´t understad wath is ???. The boinc help does not give me any clue.
Why i need to do that? E@H uses a lot of bandwitdh, if you use to much bandwith on a free proxy you will kick-out/banish very fast of it.
Any help is most welcomed. Thanks in advance.
And sorry i knew this is not the right thread but can´t find any better one to ask.
The BOINC manager has the following menus:
File Close Window: Minimize the BOINC Manager to the notification area or Dock. Note that closing the manager using the [X] button on the title bar will close the window, not exit BOINC.
Exit: Exit the BOINC manager and all running BOINC applications. No further work will take place until you run the BOINC manager again. On Mac OS X, this is under the BOINC menu as Quit BOINC.
View Shows shortcut keys to Notices, Projects, Tasks, Transfers, Statistics, Disk Usage and Simple View.
Tools Add project or account manager: Enroll this computer in a project or account manager. For projects, you will be asked to enter the project's URL and your account details, which is the email address and password you desire to use for an existing or new project registration. For World Community Grid enter the member name and password previously established when registering there (required!). Also see topic Choosing and joining projects for more registration information. For account manager, you can attach to one or more new projects using an account manager web site. See Account managers.
Display and network options: Opens a dialog allowing you to select your preferred language, how often you wish to be reminded of new notices coming in, etc. If you connect to the web through an HTTP or SOCKS proxy, use this dialog to enter its address and port. You may also enter a comma-separated list of URLs for which a proxy should not be used. Windows only: use this dialog to tell BOINC your method of connecting to the Internet.
Computing preferences: Opens a dialog allowing you to override your preferences locally.
or working directly in the cc_config proxy configuration section, you obviously want to use the no_proxy optional line, specified this way in the documentation:
list of hostnames for which proxy not used
The sticky bit is how to populate the comma-delimited list of URLs, which I imagine is the question you are posing.
I think the byte-dominant traffic is work download for BRP CUDA work. It has been said here that work is served from Hannover.
Browsing through client_state.xml, and hoping that what is shown there is relevant, it seems you might want to try using:
einstein-dl.aei.uni-hannover.de
in the list. But I don't know whether it should be that way, or perhaps more fully specified in proper URL manner as:
since you are allowed a comma-delimited list, perhaps it would be well to try both.
Do you have any means to detect what your actual proxy traffic is, and thus whether an attempt configuration is successful?
This topic does not seem to get discussed much, as searches of the forums here, at SETI, and at BOINC for the full string "list of hostnames" gets no hits, and even a search using "no_proxy" does get ones obviously directly on-topic.
The BOINC manager has the following menus:
File Close Window: Minimize the BOINC Manager to the notification area or Dock. Note that closing the manager using the [X] button on the title bar will close the window, not exit BOINC.
Exit: Exit the BOINC manager and all running BOINC applications. No further work will take place until you run the BOINC manager again. On Mac OS X, this is under the BOINC menu as Quit BOINC.
View Shows shortcut keys to Notices, Projects, Tasks, Transfers, Statistics, Disk Usage and Simple View.
Tools Add project or account manager: Enroll this computer in a project or account manager. For projects, you will be asked to enter the project's URL and your account details, which is the email address and password you desire to use for an existing or new project registration. For World Community Grid enter the member name and password previously established when registering there (required!). Also see topic Choosing and joining projects for more registration information. For account manager, you can attach to one or more new projects using an account manager web site. See Account managers.
Display and network options: Opens a dialog allowing you to select your preferred language, how often you wish to be reminded of new notices coming in, etc. If you connect to the web through an HTTP or SOCKS proxy, use this dialog to enter its address and port. You may also enter a comma-separated list of URLs for which a proxy should not be used. Windows only: use this dialog to tell BOINC your method of connecting to the Internet.
Computing preferences: Opens a dialog allowing you to override your preferences locally.
Claggy
Thanks, i belive now i need to know what IP/URL E@H uses... where can i find them?
I could agree i´m not a expert, but the manual have very few information to an average user...
Thanks, i belive now i need to know what IP E@H uses... where can i find them?
After a quick look in the client_state.xml file, almost all the file references (the content between the tags that tells BOINC from where it has to download it) point to this domains:
I tink there are other mirrors not listed here, at least none of this seems to be the mirror listed in the server status page as: University of Glasgow LSC group...
Thanks Horacio, will try tomorrow with less beers on my head.
Just wonder, why that kind of info, proxy config, url, etc. are not more easy to find in the front page? Each time i try to find some info in the Boinc is a painfull work because is very dificult to easely find wath you are searching for.
RE: Leveraging, did you
)
I really don't have a satisfactory answer to your question. I understand what you want and, yes, it would be nice to have those numbers visible for the various categories of tasks when you look at the tasks list for a particular host. I also saw your suggestion for a 'pendings summary' giving the total number of pendings for a particular host. The Devs will have seen it and it could be something they would have a look at at some stage.
The problem is that the server code here is highly customised for the needs of this particular project. I think that E@H is the only project using 'Locality Scheduling' and many of the customisations to support that were developed and maintained here and this would all have to be retro-fitted each time if the project tried to keep up with the latest standard BOINC server code. As is always the case, the Dev team is small and overworked so even minor enhancements can take a while to be implemented when there are so many other fires to fight.
Personally, I use a different strategy to make sure all hosts are operating normally and for spotting any that have failed in any way. I don't even try to watch numbers of pendings, valids, invalids or errors directly. I use RAC and time of last contact. I have a lot of hosts and I know roughly what the RAC of each one should be. I know all about why RAC is a bad measure of performance (and I largely agree) but I've actually found it to be rather useful in spotting changes in performance despite this.
I watch the complete list of all my hosts which is arranged in order from 'most recent contact' to 'oldest contact'. If everything is OK, the bottom host on the list will be at a 'believable' value for last contact. If any seem to be a bit too long ago, I look at just those hosts and fix whatever the problem is (if any). Sometimes the machine may have crashed so just a reboot but quite often it's still crunching with the network connection down. Also sometimes a particular task just takes longer than usual so this blows the DCF out to the point that the host thinks it doesn't need more work for quite a while. I keep a 3.5 day cache on many hosts so a task taking 8 hours instead of 5 will cause the host to think it now has a 5.6 day cache and it wont initiate a contact for quite a while until the DCF is slowly worked down again.
I try to check the hosts list a couple of times per day but sometimes I can't and so the system isn't perfect. I'm not always physically there where the hosts are so that makes it difficult to poke and prod when necessary. The machines are largely not in air-conditioned surroundings so now that it's summer here, heat stress failures are likely to be a lot more prevalent. I've already started cleaning fans/heatsinks and replacing thermal paste where necessary.
Cheers,
Gary.
RE: I try to check the
)
Thanks, and welcome to my world, to many ships to take care, imagine i only crunch in 8 hosts to try not to explode the power bill, actualy i have more 20 light ships - with no GPUs - ready to dutty all time.
No AC of course and a very hot enviroment, Brazil = Tropical Hot/wet country, begining of Summer here to. Thanks my city is located 1000 mts high so not extrealy hot (rises to +/- 36C on hot days). But with to many hosts arroud, heat is "the problem" and some days we literaly "fry on hell" and need to shoot/slow down most of the GPU´s for a while. Or when rain to much, heavy lighting (my place is literaly over a big iron mine), so even with massive PSU - lighting is allways have the potential to make the hosts go to some "frozen wierd state"
And a very few time to babysitting. Thats way i need to find an easy way to check all the hosts (crunches or no) easely. Until them, will stay making the old hi-low game to find the number.
Bye, will not take more of your precious time, see you around
I also use the 'last contact'
)
I also use the 'last contact' method to try to keep tabs on my 9 rigs.
If one of them has not reported within a reasonable amount of time, I then have a look-see to make sure nothing has failed.
I try to find in the Boinc
)
I try to find in the Boinc Manual but i was unable to do that.
Does anybody knows how to do this?
E@H does not need a proxy, SETI needs a proxy, How i configure that?
Of course i know how to make the boinc use a proxy, i just don´t know how to make it use a proxy for the SETI project and don´t use the proxy for the E@H project.
I see something in the configuration option of the proxy: don´t use for ??? just don´t understad wath is ???. The boinc help does not give me any clue.
Why i need to do that? E@H uses a lot of bandwitdh, if you use to much bandwith on a free proxy you will kick-out/banish very fast of it.
Any help is most welcomed. Thanks in advance.
And sorry i knew this is not the right thread but can´t find any better one to ask.
On the Boinc Manager Advanced
)
On the Boinc Manager Advanced view 6 12 page (The default Boinc 7 Advanced view page is more basic):
Claggy
or working directly in the
)
or working directly in the cc_config proxy configuration section, you obviously want to use the no_proxy optional line, specified this way in the documentation:
list of hostnames for which proxy not usedThe sticky bit is how to populate the comma-delimited list of URLs, which I imagine is the question you are posing.
I think the byte-dominant traffic is work download for BRP CUDA work. It has been said here that work is served from Hannover.
Browsing through client_state.xml, and hoping that what is shown there is relevant, it seems you might want to try using:
einstein-dl.aei.uni-hannover.de
in the list. But I don't know whether it should be that way, or perhaps more fully specified in proper URL manner as:
http://einstein-dl.aei.uni-hannover.de
since you are allowed a comma-delimited list, perhaps it would be well to try both.
Do you have any means to detect what your actual proxy traffic is, and thus whether an attempt configuration is successful?
This topic does not seem to get discussed much, as searches of the forums here, at SETI, and at BOINC for the full string "list of hostnames" gets no hits, and even a search using "no_proxy" does get ones obviously directly on-topic.
Good luck, and let us know what you find.
RE: On the Boinc Manager
)
Thanks, i belive now i need to know what IP/URL E@H uses... where can i find them?
I could agree i´m not a expert, but the manual have very few information to an average user...
RE: Thanks, i belive now i
)
After a quick look in the client_state.xml file, almost all the file references (the content between the tags that tells BOINC from where it has to download it) point to this domains:
http://einstein.phys.uwm.edu
http://einstein1.aei.uni-hannover.de
http://einstein2.aei.uni-hannover.de
http://einstein-dl2.phys.uwm.edu
http://einstein-dl4.phys.uwm.edu
http://einstein.ligo.caltech.edu
I tink there are other mirrors not listed here, at least none of this seems to be the mirror listed in the server status page as: University of Glasgow LSC group...
Thanks Horacio, will try
)
Thanks Horacio, will try tomorrow with less beers on my head.
Just wonder, why that kind of info, proxy config, url, etc. are not more easy to find in the front page? Each time i try to find some info in the Boinc is a painfull work because is very dificult to easely find wath you are searching for.