[DRBD-user] DRBD9 (9.0.1) - some testing results

Chuck Smith cgasmith at comcast.net
Thu Mar 10 21:06:23 CET 2016

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


Not sure if it matters ,but the connection-mesh had mention in earlier 
drbd 9 documentation as "not fully implemented" to overcome the issue 
you declare the mesh in long form, this might give you more predictable 
results, worth a try, though dont forget to open all the relevant ports 
on your firewall.
#mesh 1-2
connection{
     host 6787-dblapro-edss port (this hosts port #) ;
     host 6788-dblapro-edss port  (this hosts port #) ;
}
#mesh 1-3
connection
{
     host 6787-dblapro-edss port (this hosts port #) ;
     host 6789-dblapro-edss port (this hosts port #) ;
}
#mesh 2-3
connection
{
     host 6788-dblapro-edss port (this hosts port #) ;
     host 6789-dblapro-edss port (this hosts port #) ;
}

On 3/10/2016 1:59 AM, Eugene Istomin wrote:
>
> Hello guys,
>
> please help us to understand strange verify behaviour described below.
>
> I have a feeling of verify emulation instead of really verify the data.
>
> Thanks!
>
> ---
>
> Best regards,
>
> Eugene Istomin
>
> On Friday, 19 February 2016 12:52:53 EET Eugene Istomin wrote:
>
> Hello,
>
> any comments? =)
>
> Verify test results are seems very strange for us.
>
> ---
>
> Best regards,
>
> Eugene Istomin
>
> On Wednesday, 17 February 2016 18:34:27 EET Eugene Istomin wrote:
>
> Hello,
>
> we done some tests with DRBD9 9.0.1
>
> Some preliminary results:
>
> #### 1. Verify depends on "connection-mesh" position. ####
>
> connection-mesh {
>
> hosts 6787-dblapro-edss 6788-dblapro-edss 6789-dblapro-edss;
>
> }
>
> If the first node is switched off - verify on the second if failed:
>
> 6788-dblapro-edss# drbdadm verify all
>
> storage: State change failed: (-15) Need a connection to start verify 
> or resync
>
> Command 'drbdsetup verify storage 0 0' terminated with exit code 11
>
> storage role:Secondary
>
> disk:UpToDate
>
> 6787-dblapro-edss connection:Connecting
>
> 6789-dblapro-edss role:Secondary
>
> peer-disk:UpToDate
>
> After "drbdadm up all" on the first (6787-dblapro-edss) and "drbdadm 
> down all" on the second (6788-dblapro-edss) seems working correctly 
> (with node 2 unavailability problems, but thats ok)
>
> storage role:Secondary
>
> disk:UpToDate
>
> 6787-dblapro-edss role:Secondary
>
> replication:VerifyS peer-disk:UpToDate done:100.00
>
> 6788-dblapro-edss connection:Connecting
>
> So, "connection-mesh" host position matters (not just for verify, may be).
>
> #### 2. Verify strange verify behaviour ####
>
> the same three nodes, start with fully OK status for all nodes.
>
> #storage role:Primary
>
> disk:UpToDate
>
> 6788-dblapro-edss role:Secondary
>
> peer-disk:UpToDate
>
> 6789-dblapro-edss role:Secondary
>
> peer-disk:UpToDate
>
> Than, the magic:
>
>  1. On first (Primary role) node # mount /dev/drbd/by-res/storage/0
>     /media/storage # dd if=/dev/urandom of=/media/storage/1 bs=1M
>     count=100 -rw-r--r-- 1 root root 104857600 Feb 17 18:17
>     /media/storage/1
>  2. Down Second node and mount as ordinary disk (we use external
>     metadata)# drbdadm down all # mount /dev/disk/by-label/storage
>     /media/storage/ # dd if=/dev/urandom of=/media/storage/2 bs=1M
>     count=100 # umount /media/storage/ # drbdadm up all # drbdadm
>     status - all UpToDate. Thats OK, no changes from DRBD metadata
>     point of view.
>  3. Second node - verify wait till sync is done
>
> # ... storage role:Secondary
>
> disk:UpToDate
>
> 6787-dblapro-edss role:Primary
>
> replication:VerifyS peer-disk:UpToDate done:100.00
>
> 6789-dblapro-edss role:Secondary
>
> replication:VerifyS peer-disk:UpToDate done:100.00
>
>  1. Mount Secondary node disk (assuming to see only "./1" file) #
>     drbdadm down all # mount /dev/disk/by-label/storage
>     /media/storage/ # ls -l /media/storage/
>
> -rw-r--r-- 1 root root 104857600 Feb 17 18:21 1
>
> -rw-r--r-- 1 root root 104857600 Feb 17 18:21 2
>
> What really verify do?
>
> #### 3. No information about speeds, timing & bandwith ####
>
> How long will replication lasts?
>
> What is current bandwidth?
>
> The questions are for initial sync. resync & verify.
>
> BTW, current implementation (for ex. done:80.94) is faulty sometimes.
>
> Thanks for DRBD9 mesh topology! =)
>
> ---
>
> Best regards,
>
> Eugene Istomin
>
>
>
> _______________________________________________
> drbd-user mailing list
> drbd-user at lists.linbit.com
> http://lists.linbit.com/mailman/listinfo/drbd-user

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20160310/87143dc6/attachment.htm>


More information about the drbd-user mailing list