rbd-nbd: restart parent process logger after forking #12222
Conversation
Fixes: http://tracker.ceph.com/issues/18070 Signed-off-by: Jason Dillaman <dillaman@redhat.com>
| @@ -493,8 +493,9 @@ static int do_map() | |||
| } | |||
|
|
|||
| if (forker.is_parent()) { | |||
| global_init_postfork_start(g_ceph_context); | |||
There was a problem hiding this comment.
maybe we can move this line out above if (forker.is_parent()) block, so it always gets called after fork()? and remove the call at line 636 in the if (g_conf->daemonize) block. so both child and parent's logger can be started, and can be stopped properly by the dtor of g_ceph_context?
There was a problem hiding this comment.
The daemon process fails if you do that -- this is what works.
There was a problem hiding this comment.
i tested the proposed using rbd-nbd map rbd, it does not crash. maybe i am missing something?
There was a problem hiding this comment.
Crashing is one thing (which should have been tested before the breaking commit was merged), but passing all the tests is another (rbd-nbd workunit and fsx in nbd mode).
There was a problem hiding this comment.
For me both versions look working. Anyway, I think the current initialization code in do_map already a bit confusing and it would be nice to try clean it up (as a separate PR, I could try this), so now any version that just fixes the issue could be ok.
There was a problem hiding this comment.
@dillaman oic! i will try to run the rbd tests next time when i am touching rbd related code.
There was a problem hiding this comment.
@tchaikov Thanks! Yes, please changes to RBD should be tested and reviewed.
rbd-nbd: restart parent process logger after forking #12222
Fixes: http://tracker.ceph.com/issues/18070
Signed-off-by: Jason Dillaman dillaman@redhat.com