Re: Best Practice
- From: Laimutis.Nedzinskas@xxxxxx
- To: ORACLE-L <oracle-l@xxxxxxxxxxxxx>
- Date: Fri, 20 Jan 2012 10:21:13 +0200
>an additional answer which is often better is a site which contains
neither
your primary nor your standby.
very correct. A small question remains how this other site has to be
isolated from primary and standby sites.
My first impression is that observer has to have conditions to access both
databases which are not BETTER than that of app servers. If observer is put
in privileged position with regard to app servers then it may fail to
observe conditions when app servers can no longer connect to primary db. On
the other hand If observer is put too far away from databases then FSF may
be compromised.
E.g., if app servers are behind a firewall but observer and databases are
not separated by firewall then firewall failure may prevent app servers
connecting to primary db but observer will not observe that.
I.e. one has to draw a diagram of hardware components involved including
network hardware (keeping in mind what end-user functionality those
hardware components serve) and then decide where the observer must go.
Btw, the third site gonna cost. E.g. take WAN ( e.g. metropolian) and two
data centers. Then observer must go into a third DC connected by WAN to
both DC's.
If it's the same building then a separate server connected to LAN seems to
be adequate.
---------------------------------------------------------------------------------
Please consider the environment before printing this e-mail
|------------>
| From: |
|------------>
>---------------------------------------------------------------------------------------------------------------------------------------------|
|Craig Hagan <hagan@xxxxxxx>
|
>---------------------------------------------------------------------------------------------------------------------------------------------|
|------------>
| To: |
|------------>
>---------------------------------------------------------------------------------------------------------------------------------------------|
|Laimutis.Nedzinskas@xxxxxx
|
>---------------------------------------------------------------------------------------------------------------------------------------------|
|------------>
| Cc: |
|------------>
>---------------------------------------------------------------------------------------------------------------------------------------------|
|pioro1@xxxxxxxxx, ORACLE-L <oracle-l@xxxxxxxxxxxxx>,
oracle-l-bounce@xxxxxxxxxxxxx
|
>---------------------------------------------------------------------------------------------------------------------------------------------|
|------------>
| Date: |
|------------>
>---------------------------------------------------------------------------------------------------------------------------------------------|
|2012.01.19 16:41
|
>---------------------------------------------------------------------------------------------------------------------------------------------|
|------------>
| Subject: |
|------------>
>---------------------------------------------------------------------------------------------------------------------------------------------|
|Re: Best Practice
|
>---------------------------------------------------------------------------------------------------------------------------------------------|
On Thu, Jan 19, 2012 at 9:26 AM, <Laimutis.Nedzinskas@xxxxxx> wrote:
> >Marcin Przepiorowski <pioro1@xxxxxxxxx>
> >It makes sens - if you use FSFO and primary database will loose
> connection to standby database and to observer it will shutdown
> automatically preventing environment from split brain when primary DC
> will be available again.
>
> ok, FSFO takes care of split brain. A standard "shoot yourself into your
> head" method is used.
>
correct.
>
>
> >Is is possible to run observer in same place where are your end users
> ? It that case observer will see exactly same situation
> like your users.
>
> Oracle is telling you "The observer should be installed and run on a
> computer system that is separate from the primary and standby systems "
>
> The only question is from which point to observe both databases.
> IMHO, a natural answer: from your most important point. Take your most
> important functionality (e.g. authentication services) and make sure that
> observer detects when primary becomes unavailable to authentication
> purposes.
>
>
an additional answer which is often better is a site which contains neither
your primary nor your standby. This would
allow a complete datacenter failure to (if synchronized) allow for a
failover to occur. Your risk is that
the datacenter/network/rack containing your primary + observer fails, which
will prevent your standby from
taking over since the observer will not be able to contact it.
link between DS's was broken. And well, that is VERY correct - if standby
> is accessible then let the primary shoot itself down and use standby.
>
correct, however this can only occur if the observer isn't also isolated
from the standby.
One "good" thing is that the observer usually can abide a degree of latency
and slowish links, which could allow
you to locate it somewhere a bit further away (e.g. someone's cloud, etc).
-- craig
--
//www.freelists.org/webpage/oracle-l
--
//www.freelists.org/webpage/oracle-l
Other related posts:
