RE: compelling arguments for a ThirdParty Cluster

From: Oliver Jost <Oliver.Jost_at_ca.fujitsu.com>
Date: Thu, 17 Jan 2008 15:01:11 -0500
Message-ID: <DB508580748FAD4692628A58C4E653F502ED683B@MAIL01.Corp.FC.LOCAL>


Yes, with proper installation and hardware configuration all of the issues you've listed are not really that big. Check out the Oracle docs on OTN for specifics (e.g. Solaris offers IPMP on the interconnect to accommodate failover). There are mechanisms in place for data to be protected in the event of node failure.  

Oliver


From: oracle-l-bounce_at_freelists.org
[mailto:oracle-l-bounce_at_freelists.org] On Behalf Of LS Cheng Sent: Thursday, January 17, 2008 6:47 AM To: hrishys_at_yahoo.co.uk
Cc: raindoctor_at_gmail.com; Oracle-L
Subject: Re: compelling arguments for a ThirdParty Cluster

simple words

if Oracle Clusterware fencing is not reliable we would have seen already data corruptions issues

plus, RAC internally has algorithm to avoid data corruptions when cache fusion communications suffers problems

interconnect is not SPOF if you protect it with bonding or teaming

I am not sure what is your point about LUNs but I have been in many sites and asking for 5 1GB LUNs does not seem problem at all

Thanks

--

LSC On Jan 17, 2008 3:14 PM, hrishy <hrishys_at_yahoo.co.uk> wrote:

        Hi         

	I am looking it from a prespective off SLA's
	for ex the document mentions
	
	a)Establishing dependencies between clusterware
	storage and RAC databases to ensure that the shutdown
	in the correct sequence.The sun clusterware allows the
	oracle clusterware to be stopped before the
	filesystems are stopped.
	
	b)It also suggests that there is a scope for oracle
	clusterware to evict a node and yet not reboot it 
	immedaitely .The remaning nodes start recovering the
	transactions however the evicted continues to write to
	the datafiles during the period of eviction to the
	point where it falls down to the boot prompt it will 
	lead to corruption
	
	There are some other points
	Like
	a)interconnect connect being single point of failure
	b)How to provision voting disks and OCR should i ask
	the storage admins to carve out 5 1GB LUNS so i can 
	mirror them ?
	c)Would the storage admin object if i ask 5 1GB luns
	to be carved out for me as mostly they allocate LUNS
	of 64 GB 100Gb etc
	
	regards
	Hrishy
	




	     __________________________________________________________
	Sent from Yahoo! Mail - a smarter inbox http://uk.mail.yahoo.com
	--
	
	http://www.freelists.org/webpage/oracle-l
	
	
	




--

http://www.freelists.org/webpage/oracle-l Received on Thu Jan 17 2008 - 14:01:11 CST

Original text of this message