Re: Sharing disks across machines
Hi all Eric Rowell: Could you tell us why you post the same text 5 times ? ehehe Well, a little question for everybody: Why use replicate (ER or HDR) and not mirroring ??????? If you have 2 servers, 1 running, other 'waiting' for any failure of the first then why not have the same configuration (dbspaces, chunks, onconfig, etc) in both and only change IP address ? If the first fail, the second could start ... My idea is: have the principal chunk in one machine, and the mirror in other, and in the second server, have the 'mirror' chunk as principal and the principal as mirror. Think about .... Regards. R. Ferronato Eric Rowell <erowell@knology.net> wrote in message news:<c51sec$95d$1@terabinaries.xmission.com>... > What is the use of having the disk accessible to the both boxes? > > We currently have systems with Shared Disk running Informix on Solaris. > We use Legato AAM to handle Application Clustering. This is used in > case of a failure of the Sun hardware. This will allow the disk to be > attached to the second server and the database to be brought on-line. > The Disk is setup in a RAID 5 +1 to lower the chance of Disk failure. > > We also have systems using HDR and these are separated to different > locations to prevent lose of access due to a site being lost. These do > not have as much disk redundancy. > > The IDS engine cannot write to the shared disk when another system is > using. So only one instance can be running on a given set of disk. > With this in mind HDR should be setup where each has its own disks. > Using links you can setup the disks with no issues and have the disk > mount on both systems but you would still need to have separate disk for > each system. I would suggest if they are going to use HDR they not even > think about using Shared Disk. Just setup a private network for the HDR > instance to talk on. It can be a direct cable or switched network > depending on distance. Either way I would suggest having the network > setup as its own private domain. > > > Sorry to the point... > > I wouldn't say "mad" but I would ask for someone what they expect to > gain in that configuration? > > > > > On Wed, 2004-04-07 at 07:50, Andy Kent wrote: > > Client has presented me with an, erm, interesting proposal for an HDR > > pair: each machine's disks accessible to the other box, all RAID 1+0 > > but with the mirrors on the opposite box. > > > > Clever and ingenious or barking mad? I can't decide and would welcome > > any comments. > > > > Also would it actually physically work? Would IDS be able to see the > > remote disks? > > > > We're talking 9.40 on new Sun kit. > > > > REAL, LIVE email reply address. > > > > Andy Kent > > Bristol, UK > > > > sending to informix-list > sending to informix-list > sending to informix-list > sending to informix-list > sending to informix-list
Ferronato wrote: > [...] Why use replicate (ER or HDR) and not mirroring ??????? ER - because you can't usually have 20 machines all accessing the same disk system. HDR - because if one machine falls over, the other can continue. > If you have 2 servers, 1 running, other 'waiting' for any failure of > the first then why not have the same configuration (dbspaces, chunks, > onconfig, etc) in both and only change IP address ? > If the first fail, the second could start ... That's more or less what HDR does. It's a bit better than that - it allows you to use the secondary for queries - so it is available for use as long as you don't try to modify the data. And it is kept in sync. The i/p address arrangements depend mostly on the rest of the high availability solution. > My idea is: have the principal chunk in one machine, and the mirror in > other, and in the second server, have the 'mirror' chunk as principal > and the principal as mirror. Seems a bit unnecessarily complicated, but not by very much. The main issue is the start-up time because the secondary must be offline and not accessing the disks until the primary crashes. HDR avoids those problems. -- Jonathan Leffler #include <disclaimer.h> Email: jleffler@earthlink.net, jleffler@us.ibm.com Guardian of DBD::Informix v2003.04 -- http://dbi.perl.org/
Why? I thought you said elsewhere in the thread that both machines *could* access the disks simultaneously. (Obviously, preferably not the same disks and definitely not the same bit of the same disks...) Jonathan Leffler <jleffler@earthlink.net> wrote in message news:<2aKec.7648$A_4.6070@newsread1.news.pas.earthlink.net>... > The main > issue is the start-up time because the secondary must be offline and > not accessing the disks until the primary crashes. HDR avoids those > problems.
Why? I thought you said elsewhere in the thread that the two IDS servers *could* access one another's disks simultaneously (subject to the obvious caveat emptors about makign sure they don't go anywhere near the same bit of disk) Jonathan Leffler <jleffler@earthlink.net> wrote in message news:<2aKec.7648$A_4.6070@newsread1.news.pas.earthlink.net>... > The main > issue is the start-up time because the secondary must be offline and > not accessing the disks until the primary crashes. HDR avoids those > problems.