[DRBD-user] drbdadm verify status starts at 100% and stays there?

[jj at lp.pl]

On Sun, 14 Jan 2018 17:36:39 +0100, Christiaan den Besten wrote:
>
> On 04/08/17 6:51, Eric Robinson wrote:
>> (...)
>> …which looks like the verify is still in progress.
>>
>>  
>>
>> So is it done, or is it still in progress? Is this a drbd bug?
>
> I see the same behaviour still on a fresh 9.0.11 install and latest 
> tools.
>
> r_drbd9.prolocation.net role:Secondary
>   volume:0 disk:UpToDate
>   volume:1 disk:UpToDate
>   mhxen20.prolocation.net role:Primary
>     volume:0 replication:VerifyS peer-disk:UpToDate done:100.00
>     volume:1 replication:VerifyS peer-disk:UpToDate done:100.00
>
> En remains in that state. Even if I unmount/remount it on either host 
> of
> the 2 node setup.

Same problem here - drbdadm verify often never finishes when there is 
ongoing i/o to the volume being checked.
Installed versions: debian9, drbd-dkms_9.0.11-1_all.deb, 
drbd-utils_9.2.0-1_amd64.deb

/etc/drbdmanaged.cfg:
[GLOBAL]
storage-plugin = drbdmanage.storage.lvm.Lvm

/var/lib/drbd.d/drbdmanage_global_common.conf:
common {
    disk {
        c-fill-target 2048;
        al-extents 1237;
        resync-rate 33000;
        c-plan-ahead 5;
        c-min-rate 10240;
        disk-flushes no;
        md-flushes yes;
        disk-barrier no;
    }
    net {
        verify-alg md5;
        max-epoch-size 8192;
        max-buffers 8192;
    }
}

Disabling the dynamic syncer (c-plan-ahead=0) does not help.
Re-creating fresh resource metadata on all 3 nodes does not help.

One workaround I found working: call drbdsetup verify directly, and 
specify the --stop <sector> matching drbd device size:
the problem of "never finishing" goes away, but now the verify 
sometimes finishes early (again this seems related do ongoing i/o).
In such case dmesg says "Online Verify reached sector <nnn>" and i need 
to re-run drbdsetup verify --stop <sector> again to reach the end of 
device.

--
   JJ