Thread: using warm standby with drbd???

using warm standby with drbd???

From
"libra dba"
Date:
Hi All!

I have configured the warm standby using WAL FILES and DRBD. The wal files are generated as server A. I am using DRBD to replicate these wal files between server A and server B. The DRBD service is running as Primary on Server A and secondary on Server B. The wal_files are written to the server B, but there is a drawback for DRBD. The filesystme needs to be unmounted on the server B (where DRBD is running secondary role).

So recovery is not countinuous.(file systme is unmounted). To start the recovery, i need to change the DRBD role to primary on server B. After that i have to mount the filesystem. Once i mount the filesystem, the recovery process starts.

Is there any way in which i need not switch secondary/primary role for DRBD on server B?
Is there any way in which i can have the file system mounted on server B, running DRBD secondary role?


How else can i replicate the wal_files? ( i don't want to user common file system ,,, NFS,,, etc.)?

Another thing which i want to ask is that if we are generating archives every 1 minute. then what happens to the data which was written to the server A after 35 seconds after the last wal file generation.(server A crashes).

Since this data has not been archived (it was supposed to archive after 1 minute), do i think that this is the loss of data in a server A crash as this 35 second data has not been written to the archived logs and neither has been transported to server B?

A quick reply is highly appreciated!

Thanks!

Re: using warm standby with drbd???

From
Alan Hodgson
Date:
On Tuesday 04 March 2008, "libra dba" <libra.dba@gmail.com> wrote:
> How else can i replicate the wal_files? ( i don't want to user common
> file system ,,, NFS,,, etc.)?

scp

>
> Another thing which i want to ask is that if we are generating archives
> every 1 minute. then what happens to the data which was written to the
> server A after 35 seconds after the last wal file generation.(server A
> crashes).
>

It's gone.

If that isn't acceptable then I would suggest putting the active pg_xlog
directory on drbd. That may have speed implications.

--
Alan

Re: using warm standby with drbd???

From
Erik Jones
Date:
On Mar 4, 2008, at 6:02 PM, Alan Hodgson wrote:

> On Tuesday 04 March 2008, "libra dba" <libra.dba@gmail.com> wrote:
>> How else can i replicate the wal_files? ( i don't want to user common
>> file system ,,, NFS,,, etc.)?
>
> scp

Actually, scp is a bad choice for transfering wal files if you're
planning on continuous replay.  You want something that supports
atomic transfers.  rsync is a typical choice and I've been meaning to
check out unison for a while now.

Erik Jones

DBA | Emma®
erik@myemma.com
800.595.4401 or 615.292.5888
615.292.0777 (fax)

Emma helps organizations everywhere communicate & market in style.
Visit us online at http://www.myemma.com