Re: moving instance old drives to new

Kevin wrote in message <7h9uar$90q$1@ffx2nh3.news.uu.net>... > >Q. Is this the way to go? Am I missing a far easier scenario? If you have used symbolic links for your chunk names, as strongly recommedned in TFM, you sould be able to use dd to copy the data from the old disk areas to the new, then break the links, re-create them to the new ones, bring up the server and away you go. In fact, an arguably easier way to do this (certainly fewer steps to get wrong) would be to do an ontape archive to tape, stop the engine, break and re-create the links, then use ontape to restore to the new disk areas. If you haven't used links for chunk names you can, with the help of your SysAdmin, try to create the new disks with the same controller/target IDs as the old, but from experience with Solaris 2.5 it's a nightmare and highly susceptible to cock-up. 100GBytes of data would probably have an export size of 20-30GBytes, so you might have the space. Onunload/onload is another option. >Q. Can the INFORMIXSERVER and aliases be renamed back to the old names once >the original instance is dropped? Yes. >Q. What are the gotchas for creating a 2nd instance (temporarily) on the >same machine. Kernal parameters need to be changed? Etc. ETc. >I am planning on capping each instance to only use 1/2 the memory with the >SHMTOTAL parameter. Anything else??? Shared memory and number of semaphore kernel settings are the most likely pitfalls. Neil Truby Londos Stores Hampton Hill, UK