Monit 5.22 segfault

[Former user]

trace:

Program received signal SIGSEGV, Segmentation fault.
0x000000000042b9a9 in _checkProcessResources ()
(gdb) thr apply all bt

Thread 2 (Thread 0x7ffff6228700 (LWP 2811)):
#0  0x00007ffff6b5e84d in poll () at ../sysdeps/unix/syscall-template.S:81
#1  0x000000000043c110 in Engine_start ()
#2  0x000000000041c7e2 in thread_wrapper ()
#3  0x00007ffff74ac184 in start_thread (arg=0x7ffff6228700) at pthread_create.c:312
#4  0x00007ffff6b6bbed in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:111

Thread 1 (Thread 0x7ffff7fec740 (LWP 2803)):
#0  0x000000000042b9a9 in _checkProcessResources ()
#1  0x000000000042dbc2 in check_system ()
#2  0x000000000042d65a in validate ()
#3  0x0000000000413548 in do_action ()
#4  0x00000000004142c1 in main ()

[Former user]

Bah, opened it as anonymous user so i can't edit it now

[Tildeslash]

Please can you attach the coredump?

[Former user]

If you'll tell me how :)

[Tildeslash]

You can attach files via "More" (menu on right side) -> "Attach files"

[Former user]

No I mean how to I get the coredump in the first place. I followed your comment here: https://bitbucket.org/tildeslash/monit/issues/331/monit-517-segfaults but I can't find how to get it

[Tildeslash]

you need to enable coredumps first:

ulimit -c unlimited

then start monit in the same console ... when monit will crash, the coredump will be saved either to current directory or global coredump directory such as /var/crash/core (depends on your sysctl kernel.core_pattern settings)

[Former user]

Is this what you meant?

[Tildeslash]

yes, thanks for data

[Tildeslash]

Please can you send your monit configuration file to [e-mail]?

[Former user]

sent

[Tildeslash]

The problem is triggered by the "totalmem" statement in the context of "system" service:

check system $HOST
    ....
    if totalmem > 90% for 30 cycles then alert

The "totalmem" test is for "process" service types only (it checks the total memory usage of the process and its children). It shouldn't work in the context of the "system" check ... will fix the configuration parser.

Please replace the "totalmem" test in the "check system" context with "memory" test:

    if memory usage > 90% for 30 cycles then alert

[Former user]

It works in 5.20 :)

[Tildeslash]

in 5.20.0 it doesn't crash, but the test doesn't work too, as the "totalmem" test cannot be performed (no data in check system context)

[Tildeslash]

Fixed: Issue #614: Monit may crash if "total memory" test is used in "check system" context. The "total memory" test is for process checks only (includes children memory total) and will produce syntax error if used in system check now.

→ <<cset 692514c5f207>>