Hide Forgot
Version-Release number of selected component: postgresql-server-9.4.5-1.fc23 Additional info: reporter: libreport-2.6.3 backtrace_rating: 4 cmdline: 'postgres: checkpointer process ' '' '' '' '' '' '' '' crash_function: errfinish executable: /usr/bin/postgres global_pid: 8936 kernel: 4.2.8-libre.300.fc23.gnu.x86_64 runlevel: N 5 type: CCpp uid: 26 Truncated backtrace: Thread no. 1 (10 frames) #2 errfinish at elog.c:572 #3 UpdateControlFile at xlog.c:4659 #4 CreateCheckPoint at xlog.c:8162 #5 ShutdownXLOG at xlog.c:7943 #6 CheckpointerMain at checkpointer.c:413 #7 AuxiliaryProcessMain at bootstrap.c:432 #8 StartChildProcess at postmaster.c:5243 #9 reaper at postmaster.c:2718 #11 __select_nocancel at ../sysdeps/unix/syscall-template.S:84 #12 ServerLoop at postmaster.c:1642 Potential duplicate: bug 1167105
Created attachment 1115968 [details] File: backtrace
Created attachment 1115969 [details] File: cgroup
Created attachment 1115970 [details] File: core_backtrace
Created attachment 1115971 [details] File: dso_list
Created attachment 1115972 [details] File: environ
Created attachment 1115973 [details] File: limits
Created attachment 1115974 [details] File: maps
Created attachment 1115975 [details] File: mountinfo
Created attachment 1115976 [details] File: namespaces
Created attachment 1115977 [details] File: open_fds
Created attachment 1115978 [details] File: proc_pid_status
Created attachment 1115979 [details] File: var_log_messages
What would actually be useful here is the tail end of the postmaster log --- do you have that? The stack trace says that the failure occurred while trying to write pg_control during a database shutdown. An error in that would lead to a PANIC exit, which does abort(), so the SIGABRT in itself is neither surprising nor indicative of a bug. What would be interesting to know is why the write failed --- out of disk space? somebody mangled the permissions on the file? who knows? ... but there should be an entry in the postmaster log reporting what errno we got.
I've read the log, however the postmaster log didn't have relevant shutdown information. There was enough disk space and the issue didn't happen again so far. I'll reopen if this happens again with relevant data.