SMC Mirrors need maintenance

UNIX for Beginners Q & A
1

Good Morning,

I'm on a Solaris 9 SPARC machine, running SMC (Solaris Management Console)
There appears to be two sets of submirrors of drives 1-7. One set of submirrors says "OK", and the other set all say "Needs Maintenance" Double-clicking on those doesn't reveal any information. Can anyone provide some insight into what the issue might be and/or how to find out?

Thanks

2

Additional Info:
I dug out some old notes and ran metastat :

D17: Submirror of d7
State: Needs maintenance
Invoke: metareplace d7 c1t1d0s7

Research tells me I need to "Replace a slice" Does this mean physical drive replacement?

3

If all the down mirror slices are on the same spindle (drive) then, unfortunately, yes, it's likely that drive is down (unless it happens to be a cable or power problem).

4

I suggest to run
format
select the c1t1d0
and run a read test on it.
If successful then try to resync like suggested by metastat but with a -e (in-place replacement).
For example
metareplace -e d7 c1t1d0s7

5

OK Thanks... I do have questions before running this... It looks like its format/analyze/read- correct?

The menu says the read only test "doesn't harm SunOS" and other options say "doesn't harm data"

What's the difference? I'm concerned I'll wipe something important like software or settings, since it only promises that the OS is safe.

6

Use the read-only test!

(There are write tests that temporarily corrupt then restore data blocks, one block at a time. This can harm accessing "OS" processes. Such tests should only be run in single-user mode.)

7

Thanks- to be clear.. format/analyze/read is correct?

8

Yes!

9

Output:

             pass 0
             14086/23/304

             pass 1
             14086/23/304

Total of 0 defective blocks repaired

Is that good or bad? I'm guessing that's good?

10

"0 defective blocks" is good.
Now resync the mirror with metareplace -e ...

11

Yeah its was ambiguous to me because it is also saying 0 repaired which sounds bad..

Anyway.. Thanks! The metareplaces worked.. There were actually 4 mirrors that needed it and 2 that didn't. I wonder what caused this?

12

I guess your disk had a hickup/timeout, and only some partitions had an access that faulted, and then were fault-marked by the SVM.

SPARC machines often shipped with bad old disk- and controller-firmware. Unfortunately Sun never included their firmware patches in their recommended patch clusters. One had to manually search for them patches, and sometimes go through special installation procedure, but it was worth it.