[DRBD-user] Please Help With /dev/drbd0
f.g.haas at gmx.net
Tue Sep 13 15:13:41 CEST 2011
On 2011-09-13 15:06, Felix Frank wrote:
> On 09/13/2011 03:00 PM, Nick Khamis wrote:
>> Hello Felix,
>> Thank you so much for your response. I was struggling with DRBD at
>> first, but no I'm up an running, loving it, and moving onto mysql,
>> pacemaker sharding, read and write scale-out..
> of course, your scale-out will be independent of DRBD, right? Because I
> don't see how those mix, using ext3 at least.
I suppose Nick is referring to database scale-out using MySQL replication.
>>>> After switching the primary role from mydrbd1 to mydrbd2 and vice
>>>> versa, I was able to partition the drive on both nodes.
>>> Erm, what? I don't really see what you're implying. The steps should
>>> have been:
>>> 4. do whatever you want with /dev/drbd0 (mkfs, mount)
>> When I said partition, I really meant mkfs.
>> Before, I find myself in a corner, I thought I should ask. We are
>> thinking about using the InnoDB engine, do you see any problem with
>> this? Can you share some of your experiences please.
The InnoDB engine and its derivatives are the _only_ current MySQL
engine suitable for production use and capable of guaranteeing
transaction integrity in the face of DRBD failover.
> Not first hand, but it's supposed to work great.
> If you opt for very large InnoDB logs for performance reasons, your
> failovers may take very long as after primary crash, the standby node
> will have to perform an InnoDB recovery.
... where how long this will take, is all a question of properly tuning
your InnoDB buffer pool. It's by no means a law of nature that your
failover will take a very long time if you're running MySQL/InnoDB on DRBD.
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 262 bytes
Desc: OpenPGP digital signature
More information about the drbd-user