Question : gotcha's in upgrading by swapping system drives

Question : gotcha's in upgrading by swapping system drives

Post by Anthony Mandi » Sat, 13 Mar 1999 04:00:00

> we are going to attempt to accomplish a hardware upgrade from an
> ultra 2 to an enterprise 450 by swapping the system disk.  our
> netadmin on this task has verified this is doable.

        Netadmin? Is s/he also a sysadmin? Do they know the driver
        related differences between a UE2 and a UE450?

Quote:> all database devices are on a raid cabinet.  heres the plan:

> 1. roll in the new machine (I think we can get away with putting it next
> to one rack)

        Yeah, it has lockable castors, so you can roll it in.
        Check its dimensions. Sun always does something stupid
        with their systems. In the case of the 450 its its length.

Quote:> 2. Backup old.   (I'm going to start this before we leave)

> 3. UPGRADE THE OS ON THE OLD.  We're running Solaris 2.6.0 which
> doesn't work 100% with the e450 hardware.  I've got the latest and
> greatest 2.6.1 media ready to go.

        There is no 2.6.1. I think you mean 2.6 hardware 5/98.

Quote:> 4. Remove sybase from the startup scripts on old, then shutdown
> old.

        You would be better off leaving the old alone. You may need
        it to fall back on if anything happens.

Quote:> 5. Swap disks.

        Warning Will Robinson! Danger! Danger!

Quote:> 6. Boot new.  Check to make sure that the second SCSI interface
> comes up with the same interface number, so sybase will know where to find
> its disks.  (Where does sybase define it's disks?)

        Master is defined in the run script, everything else is
        in the master..sysdevices table. These are the physical
        device paths, so if anything changes you're hosed.

Quote:> 7. Start Sybase

        Good luck. You may need it.

Quote:> my question concerns item number 6.  right now the raid devices are of the
> /dev/rdsk/c1t0xxxx variety.  if the interface comes up as '2' then the
> devices will be of the /dev/rdsk/c2t0xxxx variety.

        That's correct. And that's your problem.

Quote:> if this happens, the plan is to update the RUN script so that the master
> device is indicated as c2t0xxxx.  then the plan is to update the sysdevices
> table so that c1t0xxxx -> c2t0xxxx on old, shutdown the database, and
> move the raid cable to the new box, boot.

> anyone try this before?

> the interface should come up as '1'.  just thinking about what to do if
> it does not.

> any other gotchas?

        I would leave the old box alone. I would install Solaris on the
        UE450 from scratch. Attach the RAID device and its controller,
        reconfigure & reboot and see what it comes up as. If the
        controller id is the same you're in luck. If not, its a bit

        Install Sybase on the new box or copy the directory over.
        Edit the interfaces file to show the new IP address (if you
        copied the IP address of the old machine, make sure you change
        its one first and reboot). Run the Sybase start script. If
        the controller ID has changed, edit the run script to point
        to the new master device and run it with the "-m" option
        (i.e. master only). Then change all the sysdevices entries
        to suit, shutdown and restart it without -m. Check each of
        the databases and make sure you do db dumps before you do
        any of the above. When you're done, upgrade the UE2.



Question : gotcha's in upgrading by swapping system drives

Post by Anthony Mandi » Sun, 14 Mar 1999 04:00:00

> > > 3. UPGRADE THE OS ON THE OLD.  We're running Solaris 2.6.0 which
> > > doesn't work 100% with the e450 hardware.  I've got the latest and
> > > greatest 2.6.1 media ready to go.

> >       There is no 2.6.1. I think you mean 2.6 hardware 5/98.

> the e450 is down at our colo.  what it has on it is:

> kiosk% uname -a
> SunOS kiosk 5.7 Generic sun4u sparc SUNW,Ultra-4

        This is Solaris 7. This isn't the same as Solaris 2.6.
        2.6 is a 32bit OS. Solaris 7 is a 64bit OS by default
        if installed on Ultra hardware.

Quote:> agree.   we ended up doing that.  our sysadmin figured out a way to get
> the RAID to come up as 1 though it initially came up as 2.

        The id is related to the controller port. Just moving things
        around on the same machine is enough to upset things.

Quote:> ok.  so we did the migration.  however, for some reason sybase refuses
> to use any of our indexes defined on tables.  ie : everything is table
> scanning !!

        OK. The version of Sybase hasn't changed, I gather. It may
        be related to the change in OS version. Does seem odd though.
        Solaris 7 does support 32bit apps flawlessly.

Quote:> i extracted a snippet and explicitely referenced the indexes that it
> should have chosen and the query flew.  but its not practical to go in
> and explicitely reference indexes on all queries in all stored procedures
> (ahem).

        Of course not.

Quote:> so i think the version of sybase we are using is incompatible with the
> OS.

        I would tend to agree. You can check on Sybase's web site if your
        version of the server is certified for Solaris 7.

Quote:> so the question is, has anyone had a problem like this?  here is
> the configuration on the 450:
>  Adaptive Server Enterprise/11.5/P/Sun_svr4/OS 5.5.1/1/OPT/Wed Sep 10 00:01:37
> PDT 1997

        This looks like ASE11.5 with no EBFs applied.

Quote:> obviously upgrading to 11.9 (or whatever the latest and greatest is) should
> be done here.  but it doesnt make any sense to me that the query optimiser
> would be table scanning _regardless_ of the fact that we are running 11.5
> on SunOS 5.7

        When your start the Sybase server, are there any messages in the
        server's errorlog that appear different to what they were before at
        start up?

        Also, its relatively simple to upgrade to 11.5.1 and apply the
        latest EBF. Running straight 11.5 on its own isn't the best.



Question : gotcha's in upgrading by swapping system drives

Post by Anthony Mandi » Mon, 15 Mar 1999 04:00:00

> since the box has 4 gig of memory, i configured the server to
> use 3 gig of that (total memory set to 1500000).

> i moved that down to under 2 gig (set memory set to 975000)
> and now sybase works as expected.   ie : indexes are chosen
> correctly by the optimizer.

> so if you running an 11.5 version of sybase, 2 gig of total
> memory is the max.

> i assume that 11.9 fixes that?

        ASE11.9.3 will, since its 64bit. ASE11.9.2 won't since
        its a 32bit version.

Quote:> is 11.5 32 bit?  so 32 bit sybase on 64 bit hardware is the
> issue?

        All server versions prior to 11.9.3 are 32bit. However,
        that isn't quite the issue. Its the addressable memory
        range of 32bits that the limit. In this case, its using
        a signed 32bit address range (how can you access negative
        memory addresses anyway?) so 2GB is the limit. Same deal
        with file sizes.



1. Swapping 'ontapes' from system to system

    Are the TAPEDEV and the TAPEBLK sizes the same on both machines ?? Also,
check the actual setting of block size at the OS level. I got this error on one
occasion, and one of the parameters that I changed was the block size at the OS
level. I am not sure if this parameter is overridden by TAPEBLK at INFORMIX
level or if the two have to be the same.


____________________Reply Separator____________________
Subject:    Swapping 'ontapes' from system to system

Date:       11/24/97 5:26 PM

I need to clone one Informix DB to another system.  Both
systems are Solaris Sparcs running Informix 7.23.
I want to take a tape made with 'ontape -s -L 0' from
machine 'A' and put it in machine 'B'.  When I try this
on machine 'B' (ontape -r) continuously prompts for:

Please mount tape 1 on /dev/rmt/1 and press Return to continue..
Restore a level 0 archive (y/n) y
Ready for level 0 tape

the system looks at the tape but contuously repeats the same three
lines as if it can not read the tape.  Both tape drives and
parameters are identical, and I can transfer 'tar' tapes
between the two systems. I just can't swap 'ontape'
archives.  Can someone please shed some light on this?
Thanks in advance.

2. Old type database

3. Phantom process on Universe

4. Gotcha's with SQL 7.0 during NT 4.0 to Win2k upgrade


6. A gotcha in upgrade program for 10.0.2

7. Error 1069 when starting SQL Server

8. Pdox 7 (NT) "gotcha's"

9. move system databases from C drive to D drive

10. Oracle File Systems Drive (Internet Drive)

11. QUESTION?: Swapping Table :QUESTION?