[DRBD-user] Recover primary node from disk failure.

Caspar Smit c.smit at truebit.nl
Mon Aug 1 13:45:52 CEST 2011

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


Hi,

To follow up on myself, there is no need to failover.
I just tested this with a demo setup and the attach command was enough
to get everything back to normal again.

Kind regards,

Caspar Smit



2011/7/29 Digimer <linux at alteeve.com>:
> On 07/29/2011 07:46 AM, Caspar Smit wrote:
>> Hi all,
>>
>> My primary node suffered a raid (md) failure and the drbd state is now:
>>
>> cs:Connected ro:Primary/Secondary ds:Diskless/UpToDate C r----
>>     ns:76247848 nr:316725148 dw:0 dr:0 al:0 bm:0 lo:0 pe:0 ua:0 ap:0
>> ep:1 wo:b oos:0
>>
>> So drbd is passing all IO to the second node.
>>
>> In the meantime I fixed the md and now i'm ready to attach the md again.
>>
>> But here's my question:
>>
>> Do i NEED a failover so the secondary node becomes primary/uptodate or
>> can drbd handle the fact that when I attach the md again the primary
>> becomes synctarget / inconsistent, will it stil pass all IO to the
>> secondary node first until the primary is uptodate again?
>>
>> in short:
>>
>> Will "drbdadm attach <resource>" be sufficient to recover from
>> Primary/Diskless or do I need to do a failover first?
>>
>> Met vriendelijke groet,
>
> Simply reconnecting should be fine.
>
> As I understand Protocol C, writes will go to both nodes, and reads will
> come from the UpToDate node. The details are probably best answered by a
> dev or Linbit person directly though, as I am somewhat guessing. I am
> sure you don't need to fail-over though.
>
> You may *want* to fail over though, if you can do so without downtime
> (ie: live-migrate a VM, is that is what is using the DRBD storage). The
> reason being that if anything happens to the communication with the
> secondary, the primary will drop to Secondary until it's local storage
> is UpToDate.
>
> --
> Digimer
> E-Mail:              digimer at alteeve.com
> Freenode handle:     digimer
> Papers and Projects: http://alteeve.com
> Node Assassin:       http://nodeassassin.org
> "At what point did we forget that the Space Shuttle was, essentially,
> a program that strapped human beings to an explosion and tried to stab
> through the sky with fire and math?"
>



More information about the drbd-user mailing list