Message boards : Number crunching : Feature request
Author | Message |
---|---|
Rebirther Send message Joined: 17 Sep 05 Posts: 116 Credit: 41,315 RAC: 0 |
I have lost some WUs during download "Transferred a partial file", some Boinc projects have the feature "resent lost result" what Iam missing here. Any updates planned for this? |
FluffyChicken Send message Joined: 1 Nov 05 Posts: 1260 Credit: 369,635 RAC: 0 |
I have lost some WUs during download "Transferred a partial file", some Boinc projects have the feature "resent lost result" what Iam missing here. Any updates planned for this? If you could point out your tasks that failed to download to you that would help, though I had a quick look and nothing indicates a problem https://boinc.bakerlab.org/rosetta/results.php?hostid=11643 I think if it task fails to download to you it retries the download OR just sends it to someone else, but normally in the second option it would fail the your task and mark it as such. You recent results list (linked above) shows everything is fine. Team mauisun.org |
Rebirther Send message Joined: 17 Sep 05 Posts: 116 Credit: 41,315 RAC: 0 |
The problem is "fetching new work" from the scheduler, marked as downloaded but not sent to me, I have seen this by other projects too in the past. |
Feet1st Send message Joined: 30 Dec 05 Posts: 1755 Credit: 4,690,520 RAC: 0 |
Those are what we've been dubbing "ghost WUs". They are fairly uncommon. And they do no harm to the project. If it were as simple as configuring a setting to resend, I'm sure they'd have done it long ago. Which projects have you seen that have such a feature? Perhaps they've added something to the standard BOINC server code. Add this signature to your EMail: Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might! https://boinc.bakerlab.org/rosetta/ |
Rebirther Send message Joined: 17 Sep 05 Posts: 116 Credit: 41,315 RAC: 0 |
Those are what we've been dubbing "ghost WUs". They are fairly uncommon. And they do no harm to the project. If it were as simple as configuring a setting to resend, I'm sure they'd have done it long ago. RCN, Neuron, Einstein and some more have "resent lost result". I think its a part of the new (standard) server code. |
dcdc Send message Joined: 3 Nov 05 Posts: 1831 Credit: 119,627,225 RAC: 11,586 |
Rebirther, do you mean results that your computer has completed and that have been lost on return? |
Rebirther Send message Joined: 17 Sep 05 Posts: 116 Credit: 41,315 RAC: 0 |
Rebirther, do you mean results that your computer has completed and that have been lost on return? No, only if my pc request new work then they are lost (ghost WUs) |
Feet1st Send message Joined: 30 Dec 05 Posts: 1755 Credit: 4,690,520 RAC: 0 |
Ah, here's a link on Einstein's forum. Appears they did this about a year and a half ago. And here's a link to the BOINC CVS discussing some updates to it. Does anyone know if there is any way for us users to determine which BOINC version R@H servers are using? And what BOINC server version brought this change? Add this signature to your EMail: Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might! https://boinc.bakerlab.org/rosetta/ |
River~~ Send message Joined: 15 Dec 05 Posts: 761 Credit: 285,578 RAC: 0 |
Does anyone know if there is any way for us users to determine which BOINC version R@H servers are using? And what BOINC server version brought this change? Don't know either of those, but the feature needed both server and client code, and the client side code came in around 4.49, iirr, which gives an idea of the vintage. If is, however, an optional feature that is switchable by each project separately, by the project admins adding a tag to the server congif file. Unless our admins/programmers have deliberately chosen to adopt it, we don't have it. It is a quite intelligent feature, and will only resend results that it feels are still needed (ie not if another computer has already returned a replacement, for example). Could I ask our prgorammers to look at this option and advise whether there is anything to prevent Rosetta using it, please? Having said which, ghost WU do not harm this project as much as they harm other projects (because we don't use redundancy, and the real pain with ghosts is the way they slow down the redundancy checks). It may be that the extra load on the db is enough to deter the team from offering it here. That issue is that every time you ask for work the server has to identify which WU it thinks you have, and compare that with the list sent to it by the client. So I won't be too suprised, and not disappointed, if the answer we get is that Rosetta has looked at this, and decided that the option would be a net disadvantage due to the impact on db throughout. Or indeed, by some issue I have not thought of :) BOINC made this an option, and an option that is disabled by deafult, for a reason. If it made sense for every project to have it, it would not have been optional. River~~ |
FluffyChicken Send message Joined: 1 Nov 05 Posts: 1260 Credit: 369,635 RAC: 0 |
Ah, here's a link on Einstein's forum. Appears they did this about a year and a half ago. It tells you on wn RPC call to the server ( I think ) and it 505 (5.0.5 ?), I know they are upto 512 (5.1.2) or something now.. Team mauisun.org |
Keck_Komputers Send message Joined: 17 Sep 05 Posts: 211 Credit: 4,246,150 RAC: 0 |
It does tell you in the messages when you do an RPC but the explanation of the version is wrong. 505 is 5.5.x, and the highest current version is 507 or 5.7.x. It probably does not make sense for this project to resend lost results. There is a database access penalty for it. This must be balanced against the DB penalty of having excess tasks on the server. Here I would expect the load of resending tasks to far outweigh the load from extra tasks since tasks are issued one at a time. BOINC WIKI BOINCing since 2002/12/8 |
FluffyChicken Send message Joined: 1 Nov 05 Posts: 1260 Credit: 369,635 RAC: 0 |
I've seen a server code come back as 512 hence why I changed my guess from 5.5.0 to 5.0.5. I also though the server code was never considered in testing if it had been checked in. The latest is always the release version. (i.e. stable is only for client side parts.) Team mauisun.org |
Keck_Komputers Send message Joined: 17 Sep 05 Posts: 211 Credit: 4,246,150 RAC: 0 |
I've seen a server code come back as 512 hence why I changed my guess from 5.5.0 to 5.0.5. I don't know why you had one return 512, the only thing I can think of is that it was one of the first versions to actually report to the host and it returned all 3 sections of the version (5.1.2). You are correct there is no stable version of the server code, only the head. It still gets tagged similarly to the lastest version of the client though, so a server pulled and compiled today would be version 5.7.2. BOINC WIKI BOINCing since 2002/12/8 |
Message boards :
Number crunching :
Feature request
©2024 University of Washington
https://www.bakerlab.org