#1845 Check that tracker import API doesn't lead to cascade of email notifications
During testing [#1766], there was lost of taskd output related to mail. It should be verified that cascade of mail for each won't be sent to the use performing the import, to users mentioned in tickets, etc. Ideally, any mail processing should be turned off for import at all - no mail expected for such operation, and it should speed up import which is rather time-consuming for large ticket volumes now.
(2011-04-12 21:00:55) psokolovsky: rcopeland: Rick, that's one ticket I wasn't sure how to do about, and one I probably won't have time to finish, but I'd like to at least discuss approach to, so you were aware/to add more info to it
(2011-04-12 21:01:32) rcopeland: ok, for the repo we just disabled the creation of notifications during a 'refresh all' operation
(2011-04-12 21:01:32) psokolovsky: rcopeland: so, the problem, I saw that importing tickets lead to stream of mail-related tasks in taskd logs
(2011-04-12 21:01:42) rcopeland: seems like we can do something similar here
(2011-04-12 21:02:47) psokolovsky: well, that was what I was thinkng - one approach to just disable email task creation, but Ialready added previo usly "don't update mode_date" flag, and that's kind of proliferation
(2011-04-12 21:03:33) psokolovsky: another approach might be to add flag/separate call which will just create ticket model, nothing more
(2011-04-12 21:03:34) rcopeland: we may end up refactoring into some kind of "don't do normal automatic processing" feature
(2011-04-12 21:03:44) rcopeland: that could also work
(2011-04-12 21:03:50) psokolovsky: then, separate call, after entire import is done, make queue reindexing of entire tracker
(2011-04-12 21:04:54) psokolovsky: that might also help with import performance - it's probably adequate for import itself, but not too bright for bulk processing - it's like 1000 tickets/hr
(2011-04-12 21:05:06) rcopeland: yeah
(2011-04-12 21:05:13) rcopeland: we can look at that
(2011-04-12 21:05:39) rcopeland: I also suspect that for import-related tasks it will be better to validate the import in the request, and then actually process it offline
(2011-04-12 21:05:46) psokolovsky: rcopeland: so well, it would be nice if someone looked into that and tested that import won't spam users on big imports
(2011-04-12 21:06:20) rcopeland: I'll make sure we don't forget about it. Congratulations on the new position, by the way!
(2011-04-12 21:06:47) Dave: this ticket is probably relevant to Rick's recent work on forum imports too
(2011-04-12 21:07:11) psokolovsky: rcopeland: I did it kinda that way, there's separate validate call, then there's import call. but real operation still may end with errors, which needs to be communicated back to user
(2011-04-12 21:07:52) rcopeland: Paul: right, but that communication can happen offline (probably via email)
(2011-04-12 21:08:20) psokolovsky: if there're wasn't requirement to do it API way, I guess it better could have been done in web UI, where user uploads import doc, and can check import status over time (we had SVN import working this way once)