Experiment for server operations check...

Message boards : News : Experiment for server operations check...
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5

AuthorMessage
Juha
Volunteer tester

Send message
Joined: 18 Jun 08
Posts: 76
Credit: 113,089
RAC: 0
Finland
Message 44345 - Posted: 19 Nov 2012, 23:59:55 UTC - in response to Message 43941.  

Eric, check validator logic once again.
Such state should never happen:
http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=4007211


That is weird. I'll have to go through the logs to see what happened.

I found the problem. The result file for the additional result (#5) doesn't exist. I don't know how that would happen.

I haven't read server code carefully enough to be sure but there might a be a small window of opportunity for a late report just as the files are being scheduled for deletion. Or it might be something else.

Anyway, I made some changes that should take care of the stuck results for good.

The first part is the same as before. If the result is invalid or can't be opened the Astropulse side of the validator lies to the BOINC side to retry the validation later. I fixed that by making sure retry is signalled only when necessary, which is when your file server is not accessible.

Previously the code couldn't tell the difference between a missing file and a missing file server. I improved the ResultFile class to do better diagnosis of the problem. The information is then carried in the ResultFileError exception to the rest of the code. In the case of missing file server the BOINC side is told to retry later, otherwise a missing file will get the result marked with Validate Error and Invalid.

Also the log messages got a bit of touch up. Instead of logging that an error occurred the code now tries to tell where and and what kind of error occurred and what caused it in the first place.

And the rest of the changes. The exception handling in the code was a very fine example of how not to do exception handling. Basically it was just emulating the traditional way of returning an error code. Cleaning that up made the code easier to follow imho.

Combining all three changes into one made the patch a bit messy but considering that all of them are to same parts of the code I don't think separating the changes would have made the patch that much easier to follow. (Or I'm just too lazy to redo it.)

The amount of changes is a bit more than last time so instead of inlining the patch I'm just going to give a link to the patch file(1). It was made with git-svn. I don't know if svn patch can handle it but it does seem to be readable to patch. You very likely need to be in astropulse directory to apply the patch.

Just in case there's some problems with the patch I packaged the changed files. End result should be the same whichever you choose to use. Link to package(2).


(1) That was a direct link. In case it doesn't work here's the patch via Google Drive UI.
(2) Same as (1). Package via UI.
ID: 44345 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 44348 - Posted: 20 Nov 2012, 15:31:34 UTC

Today I recived fresh pack of cuda22 tasks along with ~same amount of cuda23 tasks.
cuda22 almost twice slower on this host. Does it meant that this issue remains unfixed ?
ID: 44348 · Report as offensive
Richard Haselgrove
Volunteer tester

Send message
Joined: 3 Jan 07
Posts: 1451
Credit: 3,268,851
RAC: 208
United Kingdom
Message 44350 - Posted: 20 Nov 2012, 16:09:38 UTC - in response to Message 44348.  

Today I recived fresh pack of cuda22 tasks along with ~same amount of cuda23 tasks.
cuda22 almost twice slower on this host. Does it meant that this issue remains unfixed ?

Different bug. Message 44303.
ID: 44350 · Report as offensive
Previous · 1 · 2 · 3 · 4 · 5

Message boards : News : Experiment for server operations check...


 
©2020 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.