Sun Storage 7110

[Duke]

Heh, so apparently not as simple as I'd thought...

 

Apparently we're going to have to work on this bug that sjl mentioned earlier - Bug ID: 6902902 ZFS should support a default setting for user/group quotas

 

Unfortunately this bug still refers to user based quotas rather than folder based quotas, which I do not think is the best way.

 

For example: a default quota of 200MB is set for all users in a share. This share is used for home directories, meaning all the users of that share can only put 200MB each in their home directory. Now imagine a teacher is given write access to that share because they need to be able to put 50MB of 'preparatory work' in each user's folder. Once they have done 4 folders (4x50MB = 200MB) they will hit the quota limit and be unable to add any more data.

 

Now, there are workarounds for this, where the teacher could be given their own specific quota that gave them 50MB x which would let them put in all the data. However, say lots of teachers needed to be able to do this - each one would need adding manually, and given a bigger quota when they needed to add more data. Again, this could be worked around by adding the data using a script run under an administrative account that had no quota, etc. etc. My point is, all of this requires extra management that the folder-based FSRM-style quota approach doesn't.

 

Regarding the post above - apparently you cannot just email Sun and add your support to a feature request. You need to log a support case with Oracle/Sun (during which it will be confirmed you have a valid support contract) and describe the issues you have with current quotas and that you think 6902902 (even if it's how I've described it) should be given attention. I understand that this is more work than originally planned and understand most of you wouldn't be willing to go through this unless the problem affects you personally.

 

EDIT - Working on submitting my own FRE, will update later.

 

Thanks anyway,

 

Chris

Edited August 6, 2010 by Duke

[Duke]

Whoop, got there in the end with my own RFE. Apologies for the triple-post, I'll be quiet now!

 

1) DOCS - Documentation for FSRM-style quotas on S7000 (PDF)

 

2) FEATURE REQUEST REFERENCE - RFE 6975235

 

3) PROCEDURE - Call Sun Support and raise a new case (Service Request / SR) during which they will confirm you have a valid support contract > Explain you have an issue with the way quotas work on the S7000, and that you are aware of RFE 6975235 > Ask to be added to RFE 6975235 and tell them you are supporting it.

 

Cheers!

 

Chris

[teejay]

Update on Domain Joining with 2008 R2 servers:

If you are decommissioning old DC's, demote them, switch them off and delete from AD. Also go thorugh every section in your DNS and make sure any reference to them has been removed, especially in the _msdcs sections. Also check through AD mmc consoles that they are not still in as an operation master or a global catalogue. Check using dcdiag that everything is working properly.

Once this is done, go onto the S7000 web interface, in the Active Directory service select Join Workgroup to take the S7000 out of the domain. Delete the S7000 account in AD and restart the CIFS service on the S7000 and then rejoin the domain through the Active Directory service. Check that an account has been created in Active Directory. All should work fine now :-)

[john]

Update on Domain Joining with 2008 R2 servers:

If you are decommissioning old DC's, demote them, switch them off and delete from AD. Also go thorugh every section in your DNS and make sure any reference to them has been removed, especially in the _msdcs sections. Also check through AD mmc consoles that they are not still in as an operation master or a global catalogue. Check using dcdiag that everything is working properly.

Once this is done, go onto the S7000 web interface, in the Active Directory service select Join Workgroup to take the S7000 out of the domain. Delete the S7000 account in AD and restart the CIFS service on the S7000 and then rejoin the domain through the Active Directory service. Check that an account has been created in Active Directory. All should work fine now :-)

 

Also, as I found out very painfully to my cost last week (and kmount on here can confirm) make sure that you are on the VERY latest software from Oracle as that has fixed masses of 2008 R2 issues, mine just joined straight away no fuss no messing etc ONCE I had upgraded to the latest software as I was running near enough a year behind on it.

[teejay]

Yes, that does help, what lan manager compatibility mode are you running on in CIFS? Got ours on 2 at the moment, with smb signing off and also turned off on the servers, just wondering if you've got it working with LM 5 and signing.

[john]

Yes, that does help, what lan manager compatibility mode are you running on in CIFS? Got ours on 2 at the moment, with smb signing off and also turned off on the servers, just wondering if you've got it working with LM 5 and signing.

 

I think its 2 but im not at work this week to check, its probably 2 if im honest as I was trying to avoid upgrading the firmware on the unit due to past bad experiences doing so.

[kmount]

Oh we always have fun don't we John.

[john]

Oh we always have fun don't we John.

 

Yup we do don't we But as ever your support Kim was amazing and shows how much its worth using people such as yourself at Cutter for supplying it as your help was just what was needed as we tired checking a lot of things before we had to resort to the Oracle that is well I guess Oracle lol

[Duke]

Hey - is anyone using ABE (Access Based Enumeration) on the S7000?

 

An option which, when enabled, performs access-based enumeration. Access-based enumeration filters directory entries based on the credentials of the client. When the client does not have access to a file or directory, that file will be omitted from the list of entries returned to the client. This option is not enabled by default.

 

Is there any performance overhead and has anyone run into any issues with enabling it?

 

Many thanks,

Chris

[teejay]

Works fine, had no problems with it and no overhead problems. Had a few 'where have all the folders gone!' moments from staff, but they soon got used to it.

[Duke]

Awesome, cheers!

 

It should mean that if students happen to find the root share for profiles or userspaces they'll only see their own home folder. They don't have permissions to get into any of the other folders anyway, but it never hurt to hide things even more...

[Hanson]

7110

replace bad Hdd

and 5 days status replacing,

how long wait end of process?

 

and

Scrub status 0.1% done (-3259h 51m) hmmm

restart storage yesterday after install update

[Duke]

The time usually starts to drop down quite quickly once it gets going. I've had statuses that said a couple of days at the start but finished in around 12 hours.

[Hanson]

insert new HDD

the indicators LED green and yelow

and in active problems *The device storage1/HDD10 has failed or could not be opened*

 

what it means ??

[Duke]

Have you got something in the event log about it recognising the new disk being inserted? Is it an Oracle/Sun branded disk with the correct firmware?

 

Chris

[Hanson]

it's a seagate st9300603ss revision 0006

can't find something about recognising new disk, only about failded or not be opened

[Duke]

Did you get the disk through Oracle or did you buy it off-the-shelf? Many higher-end SAN/NAS devices (NetApp, I assume EMC, etc.) won't accept retail disks because they write their own custom firmware for them. Part of it is because they need the disks to act in specific ways so can't rely on a random firmware release doing what they expect, but part of it is just so they can charge you their inflated disk prices.

 

EDIT: Actually, if you have a message about it not recognising the disk, the below probably won't make much difference and is unlikely to help. It's more likely the issue above about it not being an Sun/Oracle disk.

 

I still would have expected the S7000 to at least recognise inserting the drive though. As long as you're comfortable doing it and the array has redundancy, remove the drive again, wait 30 seconds, then re-insert it. Push the tray in hard before you use the arm to close the slot.

 

Chris

Edited November 8, 2011 by Duke

[Hanson]

some screens

quite all bad?

[Duke]

Okay, it basically looks like it's saying it cannot access the new disk you've inserted. I can think of three causes:

 

1) It's not a Sun branded disk so is refusing to read it

2) The new disk is faulty and the issues it's having are unrelated to the device

3) The slot on the chassis is faulty (usually very unlikely)

 

Do you have a support contract on the S7000 with Oracle or anyone else?

 

EDIT: I've contacted someone at Oracle to see if they've got anyone in your area who can help you.

Edited November 9, 2011 by Duke

[Hebdenlad]

Okay, it basically looks like it's saying it cannot access the new disk you've inserted. I can think of three causes:

 

1) It's not a Sun branded disk so is refusing to read it

2) The new disk is faulty and the issues it's having are unrelated to the device

3) The slot on the chassis is faulty (usually very unlikely)

 

Do you have a support contract on the S7000 with Oracle or anyone else?

 

EDIT: I've contacted someone at Oracle to see if they've got anyone in your area who can help you.

 

Hi there you probably will need to contact your local Oracle support people to get this resolved. Chris (Duke) is right in what he says about firmware on the drives and storage vendors restricting third party drives being installed in their storage arrays.

 

Are you sure the old drive has failed? What messages did you get from the array prior to replacing the drive?

 

Drop me a line on here and I'll see what else I can do to help.

 

Regards,

 

Phil

[Hanson]

Hi Phil

Raid status changed on Degraded

and Device status

HDD 10 (not sure) Replace

HDD 11 Hot Spare (for undefined)

[Hebdenlad]

Hi Phil

Raid status changed on Degraded

and Device status

HDD 10 (not sure) Replace

HDD 11 Hot Spare (for undefined)

 

Which was the one you replaced?

[Hanson]

hdd 10

[Duke]

I'm guessing 'for undefined' refers to HDD 10 which it now sees as an 'undefined' disk because it's not Sun firmware branded so isn't recognised.

 

I would have thought you'd have three options:

 

1) Get an Oracle support contract and get them to replace the disk under warranty

2) Try to source a Sun branded disk with the right firmware through our channels

3) There may be some way at the command line to force it to use a non-branded disk - this is not recommended and is unsupported, although I have done it with NetApp (no idea how you'd do it with S7000 though, or even if it's possible)

 

Cheers,

Chris

[Duke]

You wouldn't lose anything by trying it. If it's already using the hot-spare then your array is working correctly and can afford to lose the unrecognised drive. Sometimes a 'failed' disk will just turn out to be a temporary bad connection between the disk tray and backplane, or it will be a SMART predicted failure rather than a completely dead disk.