[DRBD-user] Re: Manual sync.

Roy Keene rkeene at psislidell.com
Mon Oct 17 16:24:06 CEST 2005

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



 	Roy Keene
 	Planning Systems Inc

On Mon, 17 Oct 2005, Lars Ellenberg wrote:

> / 2005-10-12 18:01:53 -0500
> \ Roy Keene:
>> All,
>>
>> 	I recieved some replies to this, but I think the question has
>> 	been misunderstood.
>>
>> The backing device on the two machines were once componentns in a DRBD
>> array.  One of the machines (machine-2) was turned off and time passed
>> (several weeks) so many changes occured on machine-1.  DRBD's sync
>> method is VERY slow and mostly stalling on my VPN link, it is
>> estimated that it will take atleast 2 months for the 60GB of data to
>> sync using DRBD's sync.
>
> which is interessting.  typically drbd sync, if you allow it to, almost
> outperforms a plain netcat.
>

I'm not sure exactly why it is slow.  I would blame it on the VPN, but I 
transfer the data over the VPN as well.  On the SyncSource "nr" and "nw" 
stay at 0, while the destination has an ever increase "nr" and "nw" and 
data seems to move.. very slowly.

This is doing a 60GB sync over a slow-ish VPN (tinc in TCP mode) over a 
T1 (on each end).  My actual sync method isn't netcat, but basically a 
block-level rsync (performed with various block sizes to eliminate the 
chance of having duplicate hashes, and tested on a laptop with an 80GB 
drive, and various other files)

>> I can sync it much faster using a method outside of drbd and making
>> the backing device on both machines identical.  DRBD doesn't seem to
>> be aware of my manual sync and happily tries to continue its own sync.
>> My question is:  How can I tell DRBD that my "Inconsistent" component
>> is really Consistent ?
>
> anyways:
> take DRBD down, so there won't be changes to the data
> while your outside-of-drbd-sync is active.
> then, netcat the meta-data, too.
> if using internal meta data, that already has happened,
> and if the underlying block device is of the same size on both nodes,
> that should be it. if you use internal meta data, but your underlying
> storage is not of the same size, you have to move the meta data into its
> expected location on the target device.
> if you use external meta data, it is easy: just netcat them, too.
>
> then, if you bring up drbd again, they both see
> the same "generation" tags, and if they compare them,
> they conclude: "ok, looks like we are equal. no need to sync."

I'm using "meta-disk internal", but after a full sync (with both sides 
having unloaded the drbd module) they still decide that one side is 
consistent while the other is inconsistent.

Thanks for the help, any ideas on why it's not working ?

>
> -- 
> : Lars Ellenberg                                  Tel +43-1-8178292-0  :
> : LINBIT Information Technologies GmbH            Fax +43-1-8178292-82 :
> : Schoenbrunner Str. 244, A-1120 Vienna/Europe   http://www.linbit.com :
> __
> please use the "List-Reply" function of your email client.
> _______________________________________________
> drbd-user mailing list
> drbd-user at lists.linbit.com
> http://lists.linbit.com/mailman/listinfo/drbd-user
>



More information about the drbd-user mailing list