Hi,
I have been trying for quite some time now to track a crash in redis-server (2.8). I thought I would ask here if anyone has a clue as to what I could try next to pinpoint the cause of the problem.
The redis-server I use is in appendonly and uses ZADD and ZREM. After a lot of those operations (normally 3 days of operation) I get a segfault caused by an assert in t_zset.c around line 1350. From my understanding, ZREM as found the object to delete then assert on the fact that the delete will actually work but it does not. Is there any known problem that could cause that or is there anything I could try to find why it happens.
I have try running redis with debug traces but then the crash somehow stopped.
I also tried both jemalloc and malloc and there is no difference.
I have tried running valgrind over redis-server to see if any rogue access would write in memory that it should not but could find nothing useful.
Any help would be welcome.
Thanks,
Mathieu