On 10/20/2015 01:27 PM, Christian Beer wrote: > On 10/17/2015 07:59 AM, David Anderson wrote: >> I checked in a change that should do this. >> Please review and test if you get a chance. > I augmented this with some logging in pull request 1415 > https://github.com/BOINC/boinc/pull/1415) and test this soon. It seems that some clients try to reupload whole files even if they were successful the first time. It's like they never got the reply from the server the first time.
I see two kind of error messages. The first one is what I introduced with #1415 and this is seen if the client uploads data again using the <file_upload> tag. It then gets the success message and does not retry again. I've seen this for 6.12.34 as well as 7.6.9. Example: > 2015-10-30 07:09:25.3252 [PID=8940 ] Starting upload of > PM0064_00421_362_0_1 from XXX.XXX.XXX.18 [offset=0, nbytes=4162] > 2015-10-30 07:09:25.3253 [PID=8940 ] Ended upload of > PM0064_00421_362_0_1 from XXX.XXX.XXX.18; retval 0 > ... > 2015-10-30 07:17:45.4209 [PID=8940 ] Starting upload of > PM0064_00421_362_0_1 from XXX.XXX.XXX.18 [offset=0, nbytes=4162] > 2015-10-30 07:17:45.4209 [PID=8940 ] client tried to reupload the > read-only file upload/36e/PM0064_00421_362_0_1 > 2015-10-30 07:17:45.4209 [PID=8940 ] Ended upload of > PM0064_00421_362_0_1 from XXX.XXX.XXX.18; retval 0 The second one is when the client requests a <get_file_size> message after the file was successfully uploaded. This is a transient error right now so the client retries the get_file_size request after some time. Example: > 2015-10-30 06:17:53.7699 [PID=8942 ] handle_get_file_size(): > [LATeah1085E_80.0_10399_-5.45e-10_1_0] returning zero > ... > 2015-10-30 06:17:58.0200 [PID=8941 ] Starting upload of > LATeah1085E_80.0_10399_-5.45e-10_1_0 from XXX.XXX.XXX.197 [offset=0, > nbytes=351] > 2015-10-30 06:17:58.0201 [PID=8941 ] Ended upload of > LATeah1085E_80.0_10399_-5.45e-10_1_0 from XXX.XXX.XXX.197; retval 0 > ... > 2015-10-30 06:19:02.4851 [PID=8941 ] [CRITICAL] > handle_get_file_size(): cannot open > [LATeah1085E_80.0_10399_-5.45e-10_1_0] Permission denied (13) > 2015-10-30 06:19:02.4851 [PID=8941 ] Returning error to client > XXX.XXX.XXX.197: can't open file (transient) This second type seems to be restricted to clients 6.6.20 and 6.4.5 so far. Is handle_get_file_size still used by newer clients? One way to get the older clients to stop reuploading the file is to send a nbytes=0 reply which would trigger a success in the next file_upload request (like newer clients). Regards Christian _______________________________________________ boinc_dev mailing list boinc_dev@ssl.berkeley.edu http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev To unsubscribe, visit the above URL and (near bottom of page) enter your email address.
