metadevices: how to test metadb (how to corrupt replicas/understanding replicas)


 
Thread Tools Search this Thread
Operating Systems Solaris metadevices: how to test metadb (how to corrupt replicas/understanding replicas)
# 1  
Old 07-28-2010
metadevices: how to test metadb (how to corrupt replicas/understanding replicas)

Hi all,

I recently started exploring Solaris 10.

I am testing metadevices now.
I have been reading about the state databases here: 6.State Database (Overview) (Solaris Volume Manager Administration Guide) - Sun Microsystems

So I created 3 metadbs on 2 slices (6 in total; c1t1d0s3 c1t1d0s5).
This is the setup:
c1t1d0s3 -> 3 replicas
c1t1d0s4 -> d1
c1t1d0s5 -> 3 replicas
c1t1d0s6 -> d2
d10 -> mirror of d1 and d2

After that I mounted this as /export/home and put it in vfstab.

Now I want to test what happens when a state db gets corrupt. So I used dd if=/dev/zero of=/dev/rdsk/c1t1d0s3.

First of all: is this a good way to test this?
When I type metadb -i all state databases seem ok. Does this command represents the current state of the meta dbs? I would think not.
Is there a possibility to see the current state? Are those state databases only used when mounting a filesystem? If it is only read when mounting then what is the difference between losing half and half+1 replicas as it has both the same meaning aka not being able to mount the filesystem?
/export/home is still mounted and I can still create files on it. Whenever I unmount, I cannot remount the partition. This is normal, right?

When I reboot the system does not goes to multiuser. Which is normal as it is stated in the aforementioned link. Then the replicas are shown as corrupt.
After deleting the replicas on c1t1d0s3 I still can't mount the metadevice. After a reboot, this does work. Is this normal? (I get messages about stale replicas)
Stale database: does this mean it is "corrupt" / "invalid" /...?

The second setup I tried:
c1t1d0s3 -> 4 replicas
c1t1d0s4 -> d1
c1t1d0s5 -> 3 replicas
c1t1d0s6 -> d2
d10 -> mirror of d1 and d2

Note the 4 replicas on c1t1d0s3.

So now when I corrupt 4 metadevices, the system should panic. As it is half+1 replicas.
After doing the dd command the machine does not panic. However that is what is mentioned on the aforementioned link :s I am confused here.
What am I doing wrong?

To recover from invalid/corrupt/... state replicas: should I just remove the old state replicas and create new ones and restart the machine?

From what I have read earlier I should create 3 replicas on each drive if I have 2 drives. I wonder why 3 and not 2. If I use 3 or 2 replicas doesn't seem to matter as in both situations when I loose half the replicas I will not be able to boot/mount. Or is this because when you delete the invalid/corrupt replicas you would have 2 and there should be a minimum of 3?
EDIT: "For a system with two to four drives: put two replicas on each drive." So seems I have misread.
However I wonder what is the difference as you still need to reboot after creating new replicas on the earlier failed drive. Otherwise you get a message about stale replicas.

I hope someone can give me some answers!

Kind regards

Last edited by deadeyes; 07-29-2010 at 06:07 AM..
# 2  
Old 07-29-2010
Your state database is showing no issue reason being the below one has not executed

dd if=/dev/zero of=/dev/md/rdsk/c1t1d0s3

in above md is only used for virtual file system, and in above should receive an error message. 2nd point is for running of volume manager the minimum functioning database is needed is more then 50% +1 in your case if all the meatda file is on one partition, and if you corrupt the metadb then your entire filesystem is going to get corrupted.

you can use metadb –f for deletion and addition of state replica.
# 3  
Old 07-29-2010
Quote:
Originally Posted by kumarmani
Your state database is showing no issue reason being the below one has not executed

dd if=/dev/zero of=/dev/md/rdsk/c1t1d0s3
Sorry that was a typo. I used /dev/rdsk/c1t1d0s3. And that returned no error.

Quote:
Originally Posted by kumarmani
2nd point is for running of volume manager the minimum functioning database is needed is more then 50% +1 in your case if all the meatda file is on one partition, and if you corrupt the metadb then your entire filesystem is going to get corrupted.
Do note that the meta dbs are on 2 different slices (I wanted to "simulate" having 2 hard drives). So in the first case only 50% of the replicas can be damaged and in the second case half+1 get damaged.
Also from the docs it tells that 50% of the replicas is enough to keep it running while half+1 is necessary to get your system booting again.

Quote:
Originally Posted by kumarmani
you can use metadb -f for deletion and addition of state replica.
The tools in itself are not the problem.
I want to test how the number of available replicas affects the system.
But for deleting half+1 replicas the system did not panic as mentioned in the docs. It is almost as it doesn't matter if there are 5 or 1 replicas available as the only problem I have seen happening is when I reboot the system or unmount the filesystem. (see example below with 7 replicas and overwriting 4 of them)

Thanks for your response!
# 4  
Old 07-29-2010
I do not have system where i can test it at the moment, but sure there r Gurus here who would be at our help !!
Login or Register to Ask a Question

Previous Thread | Next Thread

10 More Discussions You Might Find Interesting

1. Solaris

Creating metadevices

Hi I am new to SVM, but I would like to configure it in one of my servers. The setup is as follows: I have one server running solaris 10 , and is connected to a NetApp via iSCSI protocol. This is how I intend to do it: Ask the NetApp admin, to provide me with 3 LUNs of 100G each, them use the... (14 Replies)
Discussion started by: fretagi
14 Replies

2. Solaris

zfs raidz2 - insufficient replicas

I lost my system volume in a power outage, but fortunately I had a dual boot and I could boot into an older opensolaris version and my raidz2 7 drive pool was still fine. I even scrubbed it, no errors. However, the older os has some smb problems so I wanted to upgrade to opensolaris11. I... (3 Replies)
Discussion started by: skk
3 Replies

3. Solaris

SVM - Metadevices are offline after changing hostname solaris x86

Hi , We are facing an issue on one of our solaris x86 server, After changing the hostname and a orderly reboot , all metadevices shows offline. please let us know the steps to restore back all metadevices to working state with this new hostname (3 Replies)
Discussion started by: karthick.sh
3 Replies

4. Solaris

Metadevices deleted

Hi Guys, We have an issue, all metadb's on system was deleted and the system was rebooted. the system is currented mounted in single-user mode . its a x86 server. the volumes under SVM is as follows. / /var and /usr please suggest Please use code tags <- click the... (1 Reply)
Discussion started by: karthick.sh
1 Replies

5. Solaris

Solaris Volume Manger - Database Replicas Question - Benefits of Increasing Default Size?

Hey all! I was hoping someone knew anything about this one... I know with Solaris Volume Manager the default Database Replica size is 8192 blocks (4MB approximately) Now I know you can increase this amount but is there any point? The reason I am asking this is that I've setup mirroring on... (2 Replies)
Discussion started by: Keepcase
2 Replies

6. Shell Programming and Scripting

Test on string containing spacewhile test 1 -eq 1 do read a $a if test $a = quitC then break fi d

This is the code: while test 1 -eq 1 do read a $a if test $a = stop then break fi done I read a command on every loop an execute it. I check if the string equals the word stop to end the loop,but it say that I gave too many arguments to test. For example echo hello. Now the... (1 Reply)
Discussion started by: Max89
1 Replies

7. Solaris

Metadevices in mirroring ?

Hi Guys. I have the follow disk mappig.... My doubt is that the filesystem root is in mirroring. I can see this in the configuration but , I dont know exactly if this in mirroring mean disk in RAID. In short: Watching the configuration...Can I said if the filesystem / is in Raid? ... (3 Replies)
Discussion started by: aggadtech08
3 Replies

8. Solaris

Problem with DiskSuite replicas

Good morning, I have Solstice disk suite installed on my server. One disk broke so I sostitute it. A replica was present on this disk. I had delete and then recreate it with commands metadb -d and metadb -a. Now when I inquire for the status of replicas I see this: stp11# metadb -i ... (2 Replies)
Discussion started by: bonovox
2 Replies

9. Solaris

insufficient metadevice database replicas ERROR

First I would like to thank this forum for assisting me in setting up my 1st sunbox. Could not have done it if it had not been for you guys and google :D I have mirrored my box and have SUCCESSFULLY tested booting from the rootdisk and rootmirror successfully. I am now looking at configuring... (2 Replies)
Discussion started by: mr_manny
2 Replies

10. UNIX for Advanced & Expert Users

SDS and replicas

Hello, We are using Solstice Disk Suite on Solaris 2.7. We want to add two striped volume with six disks. On each disk, we take a slice and we create the stripe. That I want to know : Is it necessary to add two replicas on the same slice on the new disks, as made before on the others... (1 Reply)
Discussion started by: christophe
1 Replies
Login or Register to Ask a Question

Featured Tech Videos