[DRBD-user] Strange behaviour of DRBD

Adam Goryachev mailinglists at websitemanagers.com.au
Wed Oct 5 01:38:01 CEST 2016

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.

On 05/10/16 01:58, Romain Pelissier wrote:
> Hi,
> So first, sorry if you foud this question basic but I am very new to 
> drbd. Second, maybe it's perfectly normal but in this case, I will 
> need some advices from you gyus :)
> Here is my situation:
> I use 8.4.7-1 on RHEL 6.5 (2.6.32-642.4.2.el6.x86_64) on my 2 nodes. I 
> use drbd with this scenario:
> I have an lvm partition formatted as xfs.
>  This partition contain ma application database (several Gigs).
> I want to use drbd to have a copy of the partition on the remote node.
> Here is my conf:
> cat /etc/drbd.d/backup.res
> global { usage-count no; }
> common { syncer { rate 1000M; } }
> resource rc0 {
>         protocol C;
>         startup {
>                 wfc-timeout  15;
>                 degr-wfc-timeout 60;
>         }
>         net {
>                 cram-hmac-alg sha1;
>                 shared-secret "secret2secret";
>         }
>         on esm65 {
>                 device /dev/drbd0;
>                 disk /dev/vg0/lv_arc;
>                 address <>;
>                 #meta-disk internal;
>                 #flexible-meta-disk /dev/vg0/lv_drdb_md;
>                 meta-disk /dev/vg0/lv_drdb_md;
>         }
>         on phpesmbck01 {
>                 device /dev/drbd0;
>                 disk /dev/vg0/lv_arc;
>                 address <>;
>                 #meta-disk internal;
>                 #flexible-meta-disk /dev/vg0/lv_drdb_md;
>                 meta-disk /dev/vg0/lv_drdb_md;
>         }
> }
> The first sync works fine. To test it once completed, I stop drbd on 
> thr second node, mount /dec/vg0/lv_arc on a folder and I can see the 
> folders.
> Perferct!
> But..
> For the fun of it, I have renoved a folder on the secondary node drbd 
> partition (drbd was stop at that time). then umount it and start drbd 
> process.
> I check this
> cat /proc/drbd
> version: 8.4.7-1 (api:1/proto:86-101)
> GIT-hash: 3a6a769340ef93b1ba2792c6461250790795db49 build by 
> mockbuild at Build64R6, 2016-01-12 13:27:11
>  0: cs:Connected ro:Primary/Secondary ds:UpToDate/UpToDate C r-----
>     ns:72054328 nr:0 dw:19012784 dr:59035232 al:1170 bm:0 lo:0 pe:0 
> ua:0 ap:0 ep:1 wo:f oos:0
> So, everything seems to be up to date.
> But If I stop the seocndary node, mount the lvm in a folder, the 
> folder that I have deleted earlier is not there but it exists on the 
> primary node.
> Is it normal in this case that the status is UpToDate/UpToDate? How 
> can I be sure that the 2 partition are really in sync?
Never, EVER, modify the underlying data.

If you are going to insist on doing that, then you must know what you 
are doing, and shouldn't be asking about it here.
If you really want to do this, then you might end up with a totally 
broken system (as you have discovered). You *might* be lucky and get a 
working system by
1) discard the "other" data
2) make sure the modified copy is started as primary, and then do a 
verify and resync to properly bring the secondary up to date, and 
remember that it contains invalid data until the resync is completed.
You would have to ignore the status from DRBD, you have lied to DRBD, so 
it will lie to you.

What you *should* do:
1) Switch primary
2) Mount the new primary
3) Make changes as needed
4) Switch primary
5) Mount and check


Adam Goryachev Website Managers www.websitemanagers.com.au
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20161005/ab8d66cc/attachment.htm>

More information about the drbd-user mailing list