On Wed, 27 Apr 2011, [email protected] wrote:
> Dear developers
>
> I am testing the reliability of MDS in Ceph File System.
> As we know, the default setting, there are one active MDS and one standby MDS.
> I want to test the reliability of MDS.
> Here is my testing scenario:
> As easy to understand, here I assume the active MDS as mds0 and the
> standby one is mds1.
> I write data and stop the mds0 daemon by "ceph mds stop 0" at the same time.
> Will the standby mds1 change its status to active?
> I think the system should be normal and the data should not loss even
> though there is just one MDS.
> But there are many problem I met....><
>
> (1) I want to know which mds is active and standby.
> But there are different answer when I type different commands.
>
> 1. When I type "ceph mds stat".
> It shows: (0=up: active), 1 up: standby
> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Does it means mds0
Right, the 0 means mds0.
> is active and mds1 is standby ?
> 2.When I type "ceph mds dump -o -"
> It shows: mds0 is standby and mds1 is active.
>
> My question is: Why there are different status about mds0 and mds1?
> Which one is correct?
It sounds like you named the MDS's with numeric identifiers. You should
use non-numeric names to avoid this confusion, like mds.a and mds.b. The
numeric role/rank (mds0, mds1) is assigned to cmds instances dynamically
based on who is up and needed in which role at the time.
> (2) I want to know the issue about stoping the active mds.
> If the command which "ceph mds stat" can show the active mds correctly,
> the active mds must be mds0 and the other mds1 is standby.
> So I type"ceph mds stop 0".
> It shows: ?telling mds0 192.168.200.185:6800/14819 to stop?(0)
> ^^^^^^^^^^^^^^^^^^ mds1's IP.
> 1. Why it shows the system stoping mds0 but mds1's IP ?
The 'stop' command takes the dynamic role, not the name.
Hope this clears it up!
sage
>
> 2. When I type "ceph -w", here is the log:
> ======================================
> mds e43: 1/1/1 up {0=up:active}, 1 up standby
> mds e44: 1/1/1 up {0=up:stoping}, 1 up standby
> mds e45: 1/1/1 up {0=up:replay}
> mds e46: 1/1/1 up {0=up:reconnect}
> mds e47: 1/1/1 up {0=up:rejoin}
> mds e44: 1/1/1 up {0=up:active}
> =====================================
> >From the log, which MDS does the system stop?
> My command which is"ceph mds stop 0", but the log should not be
> {1=up:active}?
> It really confuse me...
>
> 3. When I type"ceph mds dump ?o ?", here is the log:
> 4920: 192.168.200.184:6800/30000 ?0? mds0.12 up:active seq 40828
> ^^^^^^^^^^^^^^^^^^^ mds0's IP
> Why does the system leave the mds0?
>
> Please help me solving these problem, I am very confused...
> Thanks a lot~~~^^
>
>
> By the way, this is my testing environment :
> ==================================================================================
> I set 7 servers which include 3 MONs(host1 host2 host3), 2 MDSs(host4
> host5) and 2 OSDes(host6 host7).
> The version ceph 0.26 is in my system.
> ==================================================================================
> --
> Best Regards,
> Stefanie Chen
> --
> To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
>
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
