MSA Storage Disks Upgrade

[rianui]

Hi Torsten,

So here is what I have done. Remember there are two SCSI cables connecting the disk enclosure to the rp3440 server. There are four ports on the single SCSI card (A1, A2, B1 and B2)One cable has been intact on port A2 and I have not moved it at all. Through this cable the volume with the production database/application is accessible. The second cable however is what I have moved around the remaining three SCSI slots on the card. Attached are three documents. One with "sautil" output from a single cable connected(port A2), the second with both cables connected (port A1 and A2) and the third with both cables connected (port A2 and B2).
I also noticed that when I connect both cables (A1&A2) "sautil /dev/ciss4" shows all the disks and even the two unassigned disks.

Regards,

Kamau

[rianui]

Hi Torsten,

Here is what I have done. Remember the rp3440 server connects to the disk array through two SCSI cables. The server has one card with four ports (A1,A2,B1&B2). One cable on A2 has been ijntact and I haven't moved this at all. Through this cable the client can access his database/application. The second cable is what I have moved around the other three ports. Attached are three documents, sautil_with_port_A2&A1, sautil_with_port_A2&B2, sautil_with_single_cable_A2. I noticed that when both cables are connected (A2 and A1) I can see all drives, even the two unconfigured disks. See attachments for more info.

Regards,

Rianui

[Torsten.]

This meets my assumption.

Cables must be connected to A1 and A2.

---------- PHYSICAL DRIVES ----------

Location       Ch   ID   Size           Status

External       1    0    36.4 GB        OK
External       1    1    36.4 GB        OK
External       1    2    36.4 GB        OK
External       1    3    36.4 GB        UNASSIGNED
External       1    4    36.4 GB        UNASSIGNED
External       2    0    36.4 GB        OK
External       2    1    36.4 GB        OK
External       2    2    36.4 GB        OK
External       2    3    36.4 GB        OK
External       2    4    36.4 GB        OK
External       2    5    36.4 GB        OK
External       2    8    36.4 GB        OK

---------- LOGICAL DRIVE 0 ----------

RAID Level      = 1+0
Size            = 173639 MB
Stripe Size     = 16 KB
Status          = RECOVERING

Connect the cables to A1 and A2 and let the RAID rebuild.

Following steps are:

- wait until RECOVERING has finished

- replace 2 batteries on the main controller inside the server

- while the server is off, pull disks from slots 4 and 5 (the new disks)

- relocate the disks from slots 13 and 14 to slots 4 and 5

Now you have 5 disks on the left, 5 disks on the right.

Put the new disks in slots 6 and 13.

Now you have 6 disks on the left, 6 disks on the right.

Power on the server.

The controller will balance the disk. Now the mirrors are on the left and on the right.

YOu can loose each side (each cable) now without data loss.

The question remains:

What to do with the new disks?

Refer to my previous thread - how the RAID is used now (plain disk access vs. VxVM etc ...)

[rianui]

The cable that is 'always' connected is on A2. The cable disconnected was on B1 and I moved it around. It is now on A1 next to A2.

[rianui]

A little confusing...... "relocate the disks from slots 13 and 14 to slots 4 and 5"

A. The enclosure had 10 disks before I added two more (disk 0-9). After adding mine there are two empty slots remaining. Even when one cable is disconnected, eight disks show light meaning they are being accessed actively while the last four disks (including the ones I added) show no activity.

Also, I cannot change the batteries now (availability) so I'll postpone that.

Regards,

Rianui

[Torsten.]

You have 10 disks in RAID0+1, this means the data are on 5 disks, the mirror of this data is on the remaining 5 disks.

The goal is to have 5 disks on the left side of the MSA on bus A1, the other 5 disks with the mirror on the right side on bus A2.

So you are fully redundant in case of one half will fail.


Currently you must not loose the right half, because the "original" AND the mirror disks are on the same bus.

Relocate the disks ONLY while the server is down and powered off!

I hope this is clear now ...

[rianui]

Hi Torsten,

Kindly bear with me. I had shut down the system to check the spare numbers of the cache batteries. After connecting everything back and with both cables on A1 and A2 I collected the sautil info again. Just confirm and then I will proceed with the next course of action.

Regards,

Rianui

[Torsten.]

Looks like you switched the cables between A1 and A2 now:

---- SCSI DEVICE SUMMARY -----------------------------------------------------

  Location   Ch    ID    Type         Capacity   Status

  external    1     0    DISK         36.4 GB    OK
  external    1     1    DISK         36.4 GB    OK
  external    1     2    DISK         36.4 GB    OK
  external    1     3    DISK         36.4 GB    OK
  external    1     4    DISK         36.4 GB    OK
  external    1     5    DISK         36.4 GB    OK
  external    1     8    DISK         36.4 GB    OK
  external    1    15    PROCESSOR
  external    2     0    DISK         36.4 GB    OK
  external    2     1    DISK         36.4 GB    OK
  external    2     2    DISK         36.4 GB    OK
  external    2     3    DISK         36.4 GB    UNASSIGNED
  external    2     4    DISK         36.4 GB    UNASSIGNED
  external    2     7    PROCESSOR

This is not a problem.

When RECOVERING has finished, re-sort the disks as I told you (with server switched off!!!)

[rianui]

Hi Torsten,

...relocate the disks from slots 13 and 14 to slots 4 and 5....slot 13 and 14 has blanks and no disks.

I guess I'm still not getting you correctly. When RAID 1+0 is configured, assuming  two disks adjacent to each other, wouldn't I remove the second disk from each pair?

The client has locked the site until Monday :( I will run your instructions and let you know the progress as soon as I get to site.

Thanks a lot!

Rianui

[Torsten.]

Look at this:

external 1 0 DISK 36.4 GB OK
external 1 1 DISK 36.4 GB OK
external 1 2 DISK 36.4 GB OK
external 1 3 DISK 36.4 GB OK
external 1 4 DISK 36.4 GB OK
external 1 5 DISK 36.4 GB OK
external 1 8 DISK 36.4 GB OK
external 1 15 PROCESSOR
external 2 0 DISK 36.4 GB OK
external 2 1 DISK 36.4 GB OK
external 2 2 DISK 36.4 GB OK

The red disks are on one bus (A1), the blue disks are on the other (A2).

But you should have 5 "red" disks and 5 "blue" disks.

Is it more clear now?

[rianui]

Hi Torsten,

Yes I understand that Channel A1 has seven disks on it and Channel A2 has three disks. Once RECOVERING is complete I should move two disks (disk ID 5&8) to Channel A2, right. Then slot in my two new drives to the next empty slots on Channel A1 and Channel A2 (slot 6&13) respectively. This is how I understand it now.

Regards,

Rianui

[Torsten.]

Correct.

Swap the disks when the server is powered off.

[Torsten.]

Any news?

[rianui]

Hi Torsten,

Yes, I have moved two disks from slots 5&6 on the left side of the enclosure to the right side in slot 10&11. I then added my two new drives to slot 5 and slot 12. Attached are two documents. One before I did the changes and the other after the changes.

Regards,

Rianui

[Torsten.]

This looks good!


If you now can tell how do you use /dev/dsk/c4t0d0 , we can decide what to do with the 2 new disks.

(used by VxVM or as a plain disk?)

[rianui]

Should be plain disks. See the output below:

#vxiod
0 volume I/O daemons running

Is there any other way to confirm this?

[Torsten.]

The disk is obviously not under LVM or VxVM control.

So probably it is a raw volume for the database.

I have no clue if the database can handle a "growing" disk, if we just make the logical drive "bigger".

Can you ask the database folks how the drive is used?

[rianui]

Growing the existing volume would have been the best option. However there is a risk. The client does not have a backup of the data on the server....faulty backup device...complicated.

Here is what I suggest; to create a new logical volume from the new drives and mount them seperately. This way the client can move anything accupying the current volume that is not critical eg. logs to the new volume thus freeing space on the current volume.

[rianui]

In future, (when backups exist ) should I want to undo the configuration I suggested, I will need to delete that volume and expand the existing volume, right?

[Torsten.]

Sounds good.

[rianui]

Hi Torsten,

The decision was finally made :-)

The two new disks will build their own logical drive and mount it seperately instead of expanding the existing one. So here is my action plan:

1. Run command to assign both unassigned drives using raid 1+0 and a stripe size of 16kb.

External       1    5    36.4 GB        UNASSIGNED

External       2    5    36.4 GB        UNASSIGNED

#  saconfig /dev/ciss4 -R 1+0 -S 16 -p 1:5 -p 2:5

2. Scan the new volume under UX, create and mount it. Is this good?

Regards,

Rianui

[Torsten.]

Sounds good.

Once the logical drive is created, do an ioscan, then create the VG and the LVOL(s).

[rianui]

Hi Torsten,

It is done! You are just the best and I want to say THANK YOU! This was a first for me....different from the EVAs I'm used to. Thank you, thank you, thank you for your invaluable support. See the end result below.

# saconfig /dev/ciss4 -R 1+0 -S 16 -p 1:5 -p 2:5

CISS: RAID SA controller on hardware path 0/4/1/0/4/0 has detected a logical
      drive status change:
        Logical drive 1 transitioned from NOT CONFIGURED to OK
        Drive type: RAID 1+0 - Disk Mirroring

# Logical drive 1 created

# saconfig /dev/ciss4

---------- PHYSICAL DRIVES ----------

Location       Ch   ID   Size           Status

External       1    0    36.4 GB        OK
External       1    1    36.4 GB        OK
External       1    2    36.4 GB        OK
External       1    3    36.4 GB        OK
External       1    4    36.4 GB        OK
External       1    5    36.4 GB        OK
External       2    0    36.4 GB        OK
External       2    1    36.4 GB        OK
External       2    2    36.4 GB        OK
External       2    3    36.4 GB        OK
External       2    4    36.4 GB        OK
External       2    5    36.4 GB        OK

# bdf

/dev/vg01/lvbackup 34816000    3336 34540704    0% /mnt/dsk2

Best Regards,

Rianui

[Torsten.]

You are welcome!