Bug #12478

Starting a migration in a process after a migration has been cancelled does not work

Added by Nick Thomas almost 4 years ago. Updated over 3 years ago.

Status:NewStart date:2016-02-03
Priority:NormalDue date:
Assignee:-% Done:


Target version:-


Per title


#1 Updated by Nick Thomas almost 4 years ago

Neither does cancelling the second migration. All you can do is restart the process.

#2 Updated by Patrick Cherry over 3 years ago

Just tried this on an archive disc. Eventually flexnbd crashes:

flexnbd[12861]: segfault at 7ebf64080308 ip 00007f3c7abe8bcc sp 00007ebf72ffcc70 error 6 in libev.so.4.0.0[7f3c7abe4000+d000]

#3 Updated by Patrick Cherry over 3 years ago


4232307522:I:13236 0x7fa2e7fff700 src/server/mirror.c:739: Starting mirror
flexnbd: ev.c:3552: ev_io_start: Assertion `("libev: ev_io_start called with corrupted watcher", ((WL)w)->next != (WL)w)' failed.
Removing 2001:41c8:50:2::628 from bond0.715 in reset for disc.archive-1.6708
Migration OUT for disc.archive-1.6708 not completed: to 2001:41c8:50:2:3::4 30000 signal SIGABRT (6)

Also available in: Atom PDF