Re: Disaster Recovery Planning

From: Mike Hodgson (mwh_junk_at_hotmail.com)
Date: 12/23/04 

Date: Thu, 23 Dec 2004 11:20:16 +1100

Most of that is more fault tolerance than DR.

For example, a failover cluster is not a DR solution as both nodes need
to share the disk resources. Plus both nodes need to be pretty close to
each other (restricted by limitations of the IO transport protocols -
FC, SCSI - and also short latency of the heartbeat network between
nodes). If there is a real disaster (eg. a fire in your data
centre/computer room or a plane flies into the building where your
cluster is located) then you lose your production environment AND your
DR environment (and probably your job...if you're still alive). A DR
environment really needs to be shared-nothing (like log shipping to a
remote site or some 3rd party replication/geo-clustering technology, or
even standard MSSQL replication for the more budget conscious); you have
to assume you'll lose the entire production environment.

RAID will only protect disk loss (not CPU, RAM, NIC, switches and a
myriad of other potential hardware issues) and minimal loss at that; 1
disk in a RAID 5 array dies then fine, hotswap it; if 2 disks in a RAID
5 array die, or the whole drive cage or RAID controlled then you're up
the creek and need to have an outage (potentially significant (how many
people keep a spare drive cage or RAID controller handy?)) to replace
hardware).

The 3 most significant factors to consider in devising your DR strategy are:
        1) maximum acceptable service disruption (down time)
        2) maximum acceptable data loss (5min? 1hr? 1day?)
        3) budget (the all important $$$)

The solution will always be a compromise in those 3 areas.

But this is getting off the topic a bit. The original question was
about documentation for pre-devised DR plans, which is a bit of big
call. While there are some guides (as people have provided in this
thread) to get you thinking about a solution, factors to consider and
how to go about implementing/documenting it, the solution itself will
require mostly your own brain (and probably several specific questions
posted to newsgroups to overcome potential issues that may arise).

Cheers,
Mike

Daniel Joskovski wrote:
> Hi Cqlboy,
> For DR you have many options, and like in life, how much many so much music
> : ) .
> Important thing is how fast your server must be online?
> If your business cannot wait, implement failover clustering, but this is
> most expensive and redundant solution.
> Also you have option of standby SQL server whish is connected to working SQL
> server in the means of LOG shipping so if you have spare server, go for it.
> If your business can wait some time you can use some sort off RAID to
> prevent lose of data in case of disk failure and if you cannot afford RAID
> you better have good backup strategy, which depends on how often your data
> are changed.
> So you can have only Full db backups, Full and Differential backups and
> Backups off transaction log.
> I tried to enumerate what I could remember.
>
> Regards,
> Daniel
>
> "Cqlboy" <Cqlboy@discussions.microsoft.com> wrote in message
> news:8DA906E5-C877-4078-8F89-9A5A1787BE3E@microsoft.com...
>
>>I've never seen a disaster recovery plan for SQL Server. Can anyone one
>>direct me to an example. This would be immensley helpful to understanding
>>how to put one together. Thanx. -Cqlboy
>
>
>

Relevant Pages

  • Re: reconstructing a RAID5 volume taking FOREVER?!?
    ... Server is running AIX5.2, ML04. ... sisioa0 is a PCI-X Dual Channel U320 SCSI RAID Adapter. ... Reference Guide for AIX" to reconstruct the RAID disk. ...
    (comp.unix.aix)
  • Re: Disck write latency > 50ms
    ... Also how can I tell how many users are connected to the Exchange 2003 ... long IO operations against a given physical disk are taking. ... criteria from "Optimizing Storage for Exchange Server 2003" to alert; ... we are to but another RAID card for this too. ...
    (microsoft.public.exchange.admin)
  • RE: Dive configuration on new cluster
    ... individual physical disk and file share resources for user and department ... Make sure you have enough nics for Public and Private networks. ... in each server. ... > Looking for suggestion on the best way to set up the raid levels, array, ...
    (microsoft.public.windows.server.clustering)
  • Re: RAID 0/RAID 5???
    ... disks, can offer better performance than RAID5, at the cost of disk space. ... RAID0 or RAID 1+0. ... Of course the cost in drives between these two is a factor, and for a server ...
    (microsoft.public.windows.server.general)
  • slackware 9.1 software raid problem
    ... Setting up a RAID system with Slackware 8 is not extremely difficult once ... mirroring the root partition and booting from that mirror was not possible. ... Each disk is attached to a different IDE chain on the motherboard. ... The ability to boot from the Slackware 8 install CD. ...
    (alt.os.linux)