Re: Mount and umount ceph volume

[Oscar Segarra]

Hi Paul,

The problem is not starting or stopping ceph server modules. I have the problem in the client side where I want to be able to poweroff my client machine even when the cephfs servers is not available. 

Thanks a lot 

El vie., 8 mar. 2019 1:49, Paul Theodoropoulos <address@hidden> escribió:

I've zero experience with ceph, however -

What about just incorporating ceph's status-checking facilities as the trigger, instead of watching the mount? for example

monit monitor:

check program ceph-status with path /usr/local/bin/ceph-status.sh
start program = "/bin/systemctl start ceph.target"
stop  program = "/bin/systemctl stop ceph\*.service ceph\*.target"
if status != 0 then restart

ceph-status.sh:

#!/bin/bash
ceph status >/dev/null 2>&1

As I said, no experience with ceph, just had a quick look at some of the documentation - I could be completely wrong about the feasability of this...

On 3/7/19 15:46, Oscar Segarra wrote:

Hi Martin, 

Thanks a lot for your quick response.

I have been making some tests but it looks your approach does not work at all:

This is my simple configuration:

cat << EOT > /etc/monit.d/monit_vdicube

check filesystem check_cephfs with path /mnt/vdicube_ceph_fs

    start program  = "/bin/mount -t ceph -o name=admin -o secret=AQDenzBcEyQ8BBAABQjoGn3DTnKN2v5hZm7gMw== 192.168.100.104:6789,192.168.100.105:6789,192.168.100.106:6789:/ /mnt/vdicube_ceph_fs"

    stop program   = "/bin/umount -f -l /mnt/vdicube_ceph_fs"

    IF CHANGED FSFLAGS THEN start

EOT

In this case when ceph monitors servers (192.168.100.104:6789,192.168.100.105:6789,192.168.100.106:6789) everything works fine. Start, stop, restart works great.

Nevertheless, If I loose connectivity with ceph servers (I manually stop them) the monit service doesn't find out and continues showing "Ok" when, of course, none of the internal data can be acceeded. This can be normal because the mount instruction is still there:

address@hidden mnt]# mount | grep ceph

192.168.100.104:6789,192.168.100.105:6789,192.168.100.106:6789:/ on /mnt/vdicube_ceph_fs type ceph (rw,relatime,name=admin,secret=<hidden>,acl,wsize=16777216)

In this scenario, If I execute the stop command as root manually from command line:

/bin/umount -f -l /mnt/vdicube_ceph_fs 

It umounts de FS immediately, however, If I stop it using the monit CLI: 

address@hidden /]# monit stop check_cephfs

address@hidden /]# monit status

Monit 5.25.1 uptime: 4m

Filesystem 'check_cephfs'

  status                       OK - stop pending

  monitoring status            Monitored

  monitoring mode              active

  on reboot                    start

  filesystem type              ceph

  filesystem flags             rw,relatime,name=admin,secret=<hidden>,acl,wsize=16777216

  permission                   755

  uid                          27

  gid                          27

  block size                   4 MB

  space total                  228 MB (of which 0.0% is reserved for root user)

  space free for non superuser 64 MB [28.1%]

  space free total             64 MB [28.1%]

  inodes total                 165

  inodes free                  -1 [-0.6%]

  data collected               Fri, 08 Mar 2019 00:35:28

System 'vdicnode04'

  status                       OK

  monitoring status            Monitored

  monitoring mode              active

  on reboot                    start

  load average                 [0.02] [0.20] [0.21]

  cpu                          1.2%us 1.0%sy 0.0%wa

  memory usage                 514.8 MB [28.3%]

  swap usage                   0 B [0.0%]

  uptime                       59m

  boot time                    Thu, 07 Mar 2019 23:40:21

  data collected               Fri, 08 Mar 2019 00:35:28

address@hidden /]#

It gets stuck  in the "stop pending" status.

In logs I can see the following:

[CET Mar  8 00:39:55] info     : 'check_cephfs' stop on user request

[CET Mar  8 00:39:55] info     : Monit daemon with PID 121791 awakened

Of course, mount is still there until I execute manually the umount command:

address@hidden /]# mount | grep ceph

192.168.100.104:6789,192.168.100.105:6789,192.168.100.106:6789:/ on /mnt/vdicube_ceph_fs type ceph (rw,relatime,name=admin,secret=<hidden>,acl,wsize=16777216)

address@hidden /]# umount -f -l /mnt/vdicube_ceph_fs

address@hidden /]# mount | grep ceph

address@hidden /]#

Even in this situation, monit status is still "stop pending"

address@hidden /]# monit status

Monit 5.25.1 uptime: 4m

Filesystem 'check_cephfs'

  status                       OK - stop pending

  monitoring status            Monitored

Any help will be welcome!

Óscar.

El jue., 7 mar. 2019 a las 22:06, address@hidden (<address@hidden>) escribió:

Hi,

we didn't test with ceph, you can try generic configuration, for example:

        check filesystem myfs with path /mydata
                start program = ...    #note: set the start command (mount)
                stop program = ...      #note: set the stop command (umount)

It is possible that monit won't be able to collect I/O statistics ... in that case we can implement support for ceph.

Best regards,
Martin


> On 7 Mar 2019, at 15:55, Oscar Segarra <address@hidden> wrote:
>
> Hi,
>
> I'd like to mount a cephfs filesystem when it is available (just checking ceph metadata server tcp port).
>
> And, on poweroff the server , i'd like to force umount the previous cephfs volume if it is already mounted. This is because if ceph metadata server is not available, the server loops infinitely trying to umount the cephfs mount point.
>
> Can theese two use cases be implemented with monit?
>
> Thanks a lot in advance
> Óscar
> --
> To unsubscribe:
> https://lists.nongnu.org/mailman/listinfo/monit-general


--
To unsubscribe:
https://lists.nongnu.org/mailman/listinfo/monit-general

-- 
Paul Theodoropoulos
www.anastrophe.com

--
To unsubscribe:
https://lists.nongnu.org/mailman/listinfo/monit-general