Comment # 4 on bug 1033098 from
Sorry for the late response.

# Grow.c
static int reshape_array()
... ...
        if (!forked && !check_env("MDADM_NO_SYSTEMCTL"))
                if (continue_via_systemd(container ?: sra->sys_name)) {
                        free(fdlist);
                        free(offsets);
                        sysfs_free(sra);
                        return 0;
                }

  both the Grow_continue() and Grow_continue_command() haven't ever invoked,
they're only working until the mdadm-grow-continue@%s.service works correctly
in back-ground.

 In this case, mdadm has sent "systemctl start mdadm-grow-continue@%s.service"
correctly, but the service got failure due to exitcode=2, and mdadm application
process has already (exited 0) at this time.

 The root cause has been locked, the ExecStart of service is used
'/sbin/mdadm', the mdadm package of Leap42.1 is too old:
# rpm -qa | grep mdadm
mdadm-3.3.1-8.1.x86_64
 The reshape can work well after I replace the /sbin/mdadm as latest source
code(v4.0, Jan-2017 Year).

  But there is one new bug about the commit of mdadm
"3a77acd7170199adc690332ded37c41f067c720e, udev-md-raid-assembly.rules: Skip
non-ready devices". in order to convenient to track this issue, I would open
one bug against SLES12 SP3, and duplicate with this bug.

  please feel free to change this bug status, or invalid or submit request
mdadm package update against "Leap42.1 maintain update".

Thanks,
-Zhilong


You are receiving this mail because: