PrimeGrid
Please visit donation page to help the project cover running costs for this month

Toggle Menu

Join PrimeGrid

Returning Participants

Community

Leader Boards

Results

Other

drummers-lowrise
1) Message boards : Proth Prime Search : Restarting/checkpointing not working on Linux? (Message 13328)
Posted 4041 days ago by Der Meister
Hi,

I recently crunched a bunch of pps_llr WUs on my Linux host. Almost all of them were valid, except the following three:
http://www.primegrid.com/orig/result.php?resultid=87160438
http://www.primegrid.com/orig/result.php?resultid=87160193
http://www.primegrid.com/orig/result.php?resultid=87160125

There is a common pattern between these failures: All these three results were interrupted due to (normal) boinc shutdown and restarted later. After that they ran to completion without any error but obviously something went wrong as the validator declared them invalid. I didn't find any valid task that has been interrupted and restarted, therfore I'm quite sure that there must be an issue whith checkpointing (does this app even checkpoint) or restarting (not from the beginning).

Any ideas?
2) Message boards : Problems and Help : Grid ready for PrimeGrid (Message 11839)
Posted 4106 days ago by Der Meister
Looks like some of your machines got the same host-id. I know there were some changes to the machanism that generates the host id and as you are running a pretty new version (6.4.1) you may already have that new code. I'm not really sure about what you can do to solve this. Maybe down-grading to an older release might help or maybe you can change the host-id manually.
3) Message boards : Developer blog : 2008 10 19 - New application version (Message 11226)
Posted 4146 days ago by Der Meister
Don't confuse the Boinc Manager with the primegrid wrapper. They are separated applications and have nothing to do with each other. And Boinc Manager indeed wastes a lot of ressources (together with the client when the manager is attached) because the xml parsing of the GUI-RPC protocol is really inefficient. There were some fixes lately that slightly improved things but I don't think they have made it into an official release yet.
4) Message boards : Number crunching : fools around again.. (Message 10895)
Posted 4169 days ago by Der Meister
I don't think so. According to this he processed nearly 7,000 PSP Sieve WUs but only discovered 2 (two!) factors in these WUs. This seems to be an extremly low ratio. The next thing is that cheating is really easy when using the sieve applications (just report that you didn't discover any factors). I strongly recommend that the WUs processed by this user are double checked and/or the user or at least this host is banned.
5) Message boards : Problems and Help : Issues caused by server upgrade, 2008 07 07 (Message 9685)
Posted 4240 days ago by Der Meister
I think this is a problem of recent BOINC server code. The user name and team name are plaintext and should not be treated as html - especially not when it comes to internal handling and storage.
Why do I think this is a problem of recent BOINC server code? Well, because it wasn't with the old one. The name of my team also contains an <AMPERSAND>. This never was a problem but a few days ago I tried to create that team on Leiden classical, which uses a recent BOINC server code. And guess what - although I entered the team name correctly and my browser submitted the page correctly (I even checked that with wireshark) everywhere on Leiden Classical and in it's stats-report the ampersand is replaced by <AMPERSAND>amp<SEMICOLON> which is not correct as these fields are plaintext fields.
6) Message boards : Number crunching : Space Family team's credit history (Message 9684)
Posted 4240 days ago by Der Meister
Some more information can be found in the archives of boinc_dev mailinglist:
http://lists.ssl.berkeley.edu/pipermail/boinc_dev/2008-July/thread.html
All the mails with "Team ownership transfer" belong to this topic.
7) Message boards : Number crunching : The Lunar Landing Challenge (Message 9657)
Posted 4241 days ago by Der Meister
Well, actually BOINC is designed to avoid that a lot of clients contact the server at once and in short intervals to get some work. The goal is simply to reduce the load on the server, especially the spikes. A server can handle a constant amount of work but if there are peaks in the workload the server might get unresponsive for some time. That is a problem every client-server-system has and therefore BOINC tries to avoid these peaks in server load. There are several mechanism for that (like exponentiall backoff in case of an error, sending multiple workunits at once, etc) and in general these mechanisms work well.

However, in case of these challenges a lot of people suddenly start up their clients and start to query the server for more and more work all at the same time. This obviously makes the server load go higher and higher as could be seen last night on the server status page. In general there are two solutions for this problem:
a) Build a server that is strong enough for even the highest peak in the load curve.
b) Try to avoid these situations completly.

Option a) is obviously quite expensive and a waste of ressources for most of the time as the server is then just running idle. And option b)... Well, I already talked about that.

Conclusion: The server problems at the beginning of a challenge like this can't really be solved. Of course, their impact can be reduced and that was already done compared to the first challenge. But the problem is not on the server's side, it's the users this time.
8) Message boards : Problems and Help : PrimeGrid Preferences (Message 9498)
Posted 4250 days ago by Der Meister
It wasn't a problem on PrimeGrid. Actually these error messages didn't have to do anything with the error. They were just generated because the real (php-) error message was printed out before the header information should be sent. But sending some html-text automatically sends the necessary header therefore modifying it is not possible any more and that triggered these error messages.

The actual error messages (which could have been found in the HTML-source code of the account page) said that fopen failed because boinc.netsoft-online.com (the page used to collect the cross project information printed on the account page) was down. Therefore the actual problem was on this page and not on primegrid.
9) Message boards : 321 Prime Search : 321 and n=5M status (Message 9393)
Posted 4260 days ago by Der Meister
Did you get the progress reported in the console (via standard output)? Because that is of no use for the wrapper. The other LLR versions write their progress into an extra file and that file is read by the wrapper application for reporting the current progress.

(By the way, it *could* be possible to change the wrapper to read the standard output of the LLR application and therefore get progress reporting working even with that linux version of LLR - if it reports its status that way. But it might get tricky.)
10) Message boards : 321 Prime Search : 321 and n=5M status (Message 9388)
Posted 4260 days ago by Der Meister
One thing to note: Linux will not have progress reporting, it's a known bug.

Why that? The Linux wrapper I worte has progress reporting. Or is it the LLR-version wich does not report it's current progress?


Next 10 posts
[Return to PrimeGrid main page]
DNS Powered by DNSEXIT.COM
Copyright © 2005 - 2020 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 2.37, 2.63, 2.77
Generated 27 Feb 2020 | 21:13:40 UTC