Re: Help! Oracle server clustering

In article <1e8276d6.0206280326.65f9cada_at_posting.google.com>, pagesflames_at_usa.net says...
>
>Pete Sharman <peter.sharman_at_oracle.com> wrote in message
>news:<afdidu0ohb_at_drn.newsguy.com>...
>> Still doesn't tell me what form of redundancy. Machine failure and disaster
>> recovery are two completely different requirements. You can't use a single
>> cluster to provide both.
>
>That's not true. You can have cluster configuration which perfectly
>protects you from machine failure and also works as disaster recovery.
>It was discussed here recently.

Not necessarily so. I mean COMPLETE disaster recovery. To do that, you must have two machines that do not share ANYTHING, including electricity grids and so on. In almost all the cases I've seen, campus clustering (or whatever term you want to use for it) is not used where nothing is shared. Try moving up to continental clustering. That may do what you want, but I still have not seen anyone do that with RAC and get acceptable performance from the interconnect traffic, particularly in an active-active scenario. In practical terms, RAC will (GENERALLY SPEAKING) not be used for disaster recovery. If you want to have the system perform as it normally does, and be truly highly available, then you will need two clusters, one remote from the other, with both clusters running RAC and either DataGuard or Advanced Replication. ENd of story as far as I'm concerned. Now, given a few years technology advancement, MAYBE continental clusters will live up to their promise, but hell, I'm an Aussie and we're born with cynicism ingrained! ;)

>Is interesting how Oracle people are insisting that RAC is for machine
>failure backup and you should use a stand-by (Data Guard) for disaster
>recovery (heard that couple of time this week).
>Which is simply not true. RAC can be a perfect part in your disaster
>recovery plans. The major difference between possibilities of RAC and
>Data Guard in terms of failure protecting, is delayed stand-by, which
>is a great feature if you don't believe your employees. Then Data
>Guard is the only option, but if you just want to be protected against
>fire, floods, Bin-Laden etc, then RAC is a sensible solution.

Agreed, RAC works in these situations where you have the separate nodes in separate buildings, but it still generally doesn't give you true HA. Take an absolute worst case scenario, where you're using RAC in this campus cluster sort of configuration. Maximum separation of the nodes at the moment is about 10 km from memory. What happens when someone detonates a nuclear bomb in that area? Both nodes are screwed, along with everything else. Are we gonna care? Probably not if we were located there, but you can see my point I hope!

Pete
>
>--
>_________________________________________
>
>Dusan Bolek, Ing.
>Oracle team leader
>
>Note: pagesflames_at_usa.net has been cancelled due to changes (maybe we
>can call it an overture to bankruptcy) on that server. I'm still using
>this email to prevent SPAM. Maybe one day I will change it and have a
>proper mail even for news, but right now I can be reached by this
>email.

HTH. Additions and corrections welcome.

Pete

SELECT standard_disclaimer, witty_remark FROM company_requirements; Received on Fri Jun 28 2002 - 11:47:06 CDT