Metadatabase questions 2

[gallows]

I remember reading some posts concerning mirroring and the metadatabase but have been unable to find them again. I think tek-tips site is having some search problems.

1. What is the recommended size for the metadatabase?

2. Our systems have /, tmp, var, and swap partitions. For me to mirror the root disk and create the meta db, I guess I can take space from swap. I think they set the swap to 4gb so I don't think this will be a problem.

3. How many should be created?

4. Most of our systems have only 2 internal hard drives and one controller. If I remember correctly, if the controller dies, obviously we are dead in the water. If the root disk dies, I should be able to boot from the mirror. Is this correct or will the mirror take over automatically?

5. Is it possible to mirror the root disk without having a meta db? The reason I ask is that looking at our current systems, I don't see any slices reserved for the meta db.

6. I can improve boot time by turning on logging for all partitions except root. Is this correct?

Thank you,
gallows

 

[bfitzmai]

Gallows:
#3 swap and tmp should be the same thing.

#4 You can automatically boot from the / mirror as follows:
Using the following command: "ls -l /dev/dsk/<root mirror>"
Example: ls -l /dev/dsk/c0t2d0s0 --> ../../devices/pci@1f,0/pci@1/scsi@4,1/sd@2,0:a

Write down the response after ../../devices

At the OK> prompt enter the following:
ok> nvalias backup_root /pci@1f,0/pci@1/scsi@4,1/sd@2,0:a

Set NVRAM setting for boot-device:
setenv boot-device disk backup_root

This way if disk fails, system will boot to backup_root.

 

[daFranze]

1) my metadb slice is about 40MB, the size increased dramatically from SDS 4.1 to 4.2

2) the easiest way is to take the 40MB from swap

3) the rule is: SDS needs at least 3 metadbs to boot; if some metadbs are unavailable you need at least a mayority (is this correct English) of 50,1% or (n/2)+1 database

4) if the controller dies the machine will be dead, it will just crash
if the root disk (or even the second disk) fails the system will continue running till next reboot; if you have 50% metadbs on disk1 and 50% on disk2 the system will not boot because of (n/2)+1 rule
you can boot Solaris 9 from CD (it contains SDS) and delete the metadbs on the failed disks (it deletes just the entry in configfiles), next boot from the disk which is ok (you can create a bootdevice just like bfitzmai pointed out)
If the system's disks are hot swapable you can replace the failed disk without stopping the system; you just format the new device, delete the references to the old metadbs, create new ones and tell SDS to enable the devices again (metareplace -e ...)

5) you can put metadbs in metadevices, but I never used this; if you think this is important I can check the documentation, or you can try yourself ->

http://docs.sun.com

6) I switch on UFS logging in /etc/vfstab for all UFS Filesystems; in case of a system crash fsck will finish much faster than without logging; in case of a normal reboot logging is not a big thing...

Best Regards, Franz
--
Solaris System Manager from Munich, Germany
I used to work for Sun Microsystems Support (EMEA) for 5 years

 

[gallows]

Thanks bfitzmai and Franz. I appreciate the response.

Franz, so you can in fact turn logging on for the root partition?
By the way, my mother was from Munich and your English is a heck of alot better than mine)

gallows

 

[daFranze]

# mount | grep logging
[b]/[/b] on /dev/md/dsk/d1 read/write/setuid/intr/largefiles/[b]logging[/b]/xattr/onerror=panic/dev=1540001 on Tue Jun 22 16:02:59 2004
/export/home on /dev/md/dsk/d3 read/write/setuid/intr/largefiles/logging/xattr/onerror=panic/dev=1540003 on Tue Jun 22 16:03:11 2004
#

so I guess yes, I never heared something contrary

Munich: the world is full of Germans, we used to be Worldchampions in travelling, I don't know if we are still #1 since economy in Germany is recessional; a lot of these travellers fall in love somewhere abroad. Where are you from?

English: tank you, your English is very good, too. Our English is worse than Nativespeakers but we can survive...

Best Regards, Franz
--
Solaris System Manager from Munich, Germany
I used to work for Sun Microsystems Support (EMEA) for 5 years

 

[bomberman]

Keep also in mind:

4) To boot from the root mirror, you have to ensure that there is a boot block on the second disk depending on how you set it up initially -

installboot /usr/platform/`uname -i`/lib/fs/ufs/bootblk /dev/rdsk/c0t1d0s0

as well as a properly working dump device -

dumpadm -d /dev/md/dsk/d2 <- or whatever naming convention you may have used for swap

Tom Pietschmann
DST Systems, Inc.
Kansas City, MO