I have another proposal for making BOINC more fault-tolerant: When a result template file is missing, the transitioner logs the event and quits. In this case one can create an empty result template file and restart the transitioner.
Would it be possible for the transitioner in such a case to behave as if an empty result template file exists, so it won't quit? -Kamran -----Original Message----- From: [email protected] [mailto:[email protected]] On Behalf Of David Anderson Sent: Tuesday, September 28, 2010 1:17 PM To: [email protected] Subject: Re: [boinc_dev] Problem with BOINC's logging I made a change so that the validator will retry every 6 hours instead of immediately in this case. Why did this happen? Did you delete an app_version? That's to be avoided - set the deprecated flag instead. -- David On 27-Sep-2010 1:24 PM, Kamran Karimi wrote: > Hi all, > > It is easy to end up with a full hard disk when a BOINC process logs a recurring error message. For example, over the weekend we had this situation because a validator process kept outputting lines similar to these: > > 2010-09-27 11:44:52.6131 [CRITICAL] [WU#3662461 iq64AT2_100702a_6_26] assign_credit_set() returned -161 > 2010-09-27 11:44:52.6141 [CRITICAL] get_pfc() [RESULT#4404635]: No AVP 495!! > 2010-09-27 11:44:52.6141 [CRITICAL] get_pfc() error: -161 > 2010-09-27 11:44:52.6141 [CRITICAL] [WU#3662461 iq64AT2_100702a_6_26] assign_credit_set() returned -161 > 2010-09-27 11:44:52.6151 [CRITICAL] get_pfc() [RESULT#4404635]: No AVP 495!! > 2010-09-27 11:44:52.6151 [CRITICAL] get_pfc() error: -161 > 2010-09-27 11:44:52.6151 [CRITICAL] [WU#3662461 iq64AT2_100702a_6_26] assign_credit_set() returned -161 > 2010-09-27 11:44:52.6169 [CRITICAL] get_pfc() [RESULT#4404635]: No AVP 495!! > 2010-09-27 11:44:52.6169 [CRITICAL] get_pfc() error: -161 > 2010-09-27 11:44:52.6169 [CRITICAL] [WU#3662461 iq64AT2_100702a_6_26] assign_credit_set() returned -161 > > By today (Monday) we had a 28GB log file and a full disk, with an inconsistent database and all sorts of other problems. > > Would it be possible to change the way BOINC processes deal with such situations? For example, in the above case it would be better if the validator would just give up on this result, or at least limit the number of log entries. > > Thanks. > > -Kamran > _______________________________________________ > boinc_dev mailing list > [email protected] > http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev > To unsubscribe, visit the above URL and > (near bottom of page) enter your email address. _______________________________________________ boinc_dev mailing list [email protected] http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev To unsubscribe, visit the above URL and (near bottom of page) enter your email address. _______________________________________________ boinc_dev mailing list [email protected] http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev To unsubscribe, visit the above URL and (near bottom of page) enter your email address.
