I don't know if this is new from Einstein or from BOINC, but it seems like a good idea to me. I had a harddrive failure on one of my systems Friday, and had to rebuild it...losing all my Seti and Einstein WUs in the process. This morning when I did an update I got the following (sorted most recent to oldest):
Host Project Date Message
Server server Einstein@Home 11/21/2005 6:00:19 AM Started download of w1_1491.0
Server server Einstein@Home 11/21/2005 6:00:19 AM Started download of w1_0266.5
Server server Einstein@Home 11/21/2005 6:00:18 AM Message from server: Resent lost result w1_1491.0__1491.4_0.1_T11_S4hC_3
Server server Einstein@Home 11/21/2005 6:00:18 AM Message from server: Resent lost result w1_1491.0__1491.1_0.1_T11_S4hC_1
Server server Einstein@Home 11/21/2005 6:00:18 AM Message from server: Didn't resend lost result w1_1491.0__1491.1_0.1_T10_S4hC_3 (expired)
Server server Einstein@Home 11/21/2005 6:00:18 AM Message from server: Resent lost result w1_1491.0__1491.3_0.1_T09_S4hC_3
Server server Einstein@Home 11/21/2005 6:00:18 AM Message from server: Resent lost result w1_0266.5__0266.9_0.1_T16_S4hC_1
Server server Einstein@Home 11/21/2005 6:00:18 AM Scheduler request to http://einstein.phys.uwm.edu/EinsteinAtHome_cgi/cgi succeeded
Note the one WU: Didn't resend lost result (expired).
Good job, whoever implemented this (BOINC 5.2.7)
Seti Classic Final Total: 11446 WU.
Copyright © 2024 Einstein@Home. All rights reserved.
Resending lost work msgs from server
)
This works because a protocol was put in place some time ago to allow the server and client parts of BOINC to sync up with each other. It was done by Bruce Allen and David Anderson (and perhaps others?) in order to solve the problem of ghost work units where the server had sent work to the client but the client had not received it.
You have now discovered a very nice side benefit in that if you lose work through a disk crash, it seems like the server will replace it when next the two sync up with each other. Did it only work for EAH and not Seti? I would have thought that all projects might be taking advantage of this.
And yes, it was a nice touch not to send expired work :).
Cheers,
Gary.
RE: You have now discovered
)
I haven't SEEN anything similar from SETI, but I haven't searched the logs completely either.
Seti Classic Final Total: 11446 WU.
Try a manual update on Seti
)
Try a manual update on Seti and see if that initiates anything. I don't remember enough of the details to be sure if it is an "EAH only" feature or if it is now there for all BOINC projects to use. Maybe something extra has to be configured on the server and the project Devs have to turn it on.
Cheers,
Gary.
RE: Try a manual update on
)
Probably not - if I recall other discussions on other boards correctly, Einstein is the only project that has put this in place on the server side so far.
Thanks for that, Bill. Did
)
Thanks for that, Bill.
Did you see any reasons why other projects haven't adopted it? I'm sure I've seen people talking about problems with ghost work units on other projects, and not just EAH.
Cheers,
Gary.
RE: Did you see any reasons
)
No... I know all projects have had the "ghost WU" problem to varying degrees. I knew about the resend feature in V5.x, and suggested someone having that problem at SETI upgrade BOINC and see if the WUs were resent - one of the devs said that only worked on Einstein so far, but gave no other details. Of course, now I don't remember the thread, the dev, or what I had for lunch...
RE: ... now I don't
)
Know the feeling well .... :).
Cheers,
Gary.
Of course, the drawback is
)
Of course, the drawback is that with the replacements, you have less time...
BOINC WIKI
RE: You have now discovered
)
Gary, when David and I added this to the BOINC scheduler, we made it an option. It's enabled on E@H but apparently not on S@H. It does make some additional database queries, so David may have it turned off to prevent additional strain on the BOINC database. E@H has a very high horsepower server for this, so it's not a problem for us.
Director, Einstein@Home