Okay :) I'd say that, after some time, renames incorrectly displaying "in progress" as status, should display "failed", and failing data recorded in logstash so devs could easily check what caused the failure. Thanks!
Description
Description
Related Objects
Related Objects
Event Timeline
Comment Actions
There is no "soft" way for rename to fail; failures will be caused by an exception, DB error or similar. That is usually recorded in Logstash (when not, or without any useful information, as in the case of some HHVM errors, there is not much we can do about it), if not always easy to find. My plan for improving discoverability is T142313 (and then maybe adding a ScopedLoggerContext which will add some flag to the logstash data of exceptions happening within its scope).
As for changing "in progress" to "failed", TBH I don't think it's worth the effort. Renames getting stuck should be very rare once ddb5f69 is in production, and all a timer would do is replicate the current human behavior of "wait a few days then report a bug" which works well enough as it is.