NetBackup 7.6.0.1 and Appliance 2.6.0.1 Install/Upgrade Field Experience

[Khajan]

We have released NetBackup 7.6.0.1 and Appliance 2.6.0.1 on 18th December 2013(Yesterday), Let us start logging our experience in this thread... I have been testing NetBackup 7.6FA for some time, soon I'll be helping few customers with 2.6.0.1 and NetBackup 7.6.0.1 soon.

For more details please check following few good links.

sort.symantec.com

entsupport.symantec.com  Enterprise NetBackup Server/NetBackup Appliances

Good Luck for your upgrades.

Khajan Gowda

[Khajan]

Part of Best Practices Symantec recommends not to combine internal 3TB with external Disk configured as MSDP pool. 

Stefan, do you have need to grow your MSDP to bigger size?

-Khajan

[StefanosM]

No at the moment, I just need to know the max limit of a deduplication pool when I have a 86 TB 5230 appliance.

[EnriquePereira]

I have upgraded mine yesterday. I had some problems:

- Be careful if you have running process that you can not kill. Sometimes there are bpbkar32 and other processes hung that are not stopped with bpdown or kill_all. The first attemt failed for this reason, but instead of stopping at the beginning, it made all the upgrade process and a rollback at the end. The rollback works really great!!! it seems that it splits the OS mirror and, if something goes wrong, it discards all the changes and resync with the original disk :) Also, I received a call from autosupport as the OS raid had a problem.

- The second problem was because I had a bad link on the databases (I made several links to solve the problem with earlier versions of having a mix of upercase/lowercase client names). I solve the problem and ran the upgrade process manually. Also, the estimation says 1,5hours for a 27TB pool and it took 5 hours, so be careful estimating times, it can take a whole day to upgrade it!

[EnriquePereira]

The limit at this moment is 72TB.

[Stanleyj]

Enrique, I seem to be having the same problem with the raid. I have two appliances and both failed during the upgrade about the raid1 config was not optimal. now when I go to rerun the update it keeps saying "Waiting for RAID 1 rebuild" Did yours say something similar. On your second issue what was the actual error? Because I to had issue with the upper/lower case client names but I cant remember if support created any links or not. I don't think we did but might have created one just as a test and then left it. I just would like to see if I get the same message then I know where to look. Thanks.

[EnriquePereira]

Stanleyj,

The mirror was rebuild automatically. It took about 3 hours. When I tried to see the status it seems that it was not rebuilding, but it did it automatically. YOU MUST WAIT FOR IT TO FINISH BEFORE TRYING AGAIN THE UPGRADE:

/opt/MegaRAID/MegaCli/MegaCli64 -ldinfo -L0 -a2

Adapter 2 -- Virtual Drive Information:

Virtual Drive: 0 (Target Id: 0)

Name                :

RAID Level          : Primary-1, Secondary-0, RAID Level Qualifier-0

Size                : 930.390 GB

State               : Degraded

Strip Size          : 64 KB

Number Of Drives    : 2

Span Depth          : 1

And the second error was:

[2014-01-07 16:18:07][DBG] Init datastore

[2014-01-07 16:18:07][INF] Processing PO list file

[2014-01-07 16:18:07][INF] Backing up catalog

[2014-01-07 16:19:54][ERR] /disk/databases/catalog/2/client/client: Failed to stat

There are errors in the conversion

Please check logs in /disk/log/convert to solve the problem and retry

But do not worry, If the link was done correctly you should not have any problems. I made an error making the link. I just removed the bad link. If you get this error, go to the directory and check if there is something wrong. When you correct it, you can rerun the upgrade with:

/usr/openv/pdde/pdconfigure/scripts/installers/PDDE_convert.sh

Good Luck !!

[Stanleyj]

Thank you so much!!

[Peter_Jakobs]

My upgrade experience was not that great.

There is no more 32bit Java administration console, this was very hard to find in the release notes.

During the upgrade of my 5220 appliance, which is only e media server, the update process updated the 

/usr/openv/db/data/vxdbms.conf file on the master server. it changed the line VXDBMS_NB_REMOTE_SERVER = 5220appliance_name

this made my database goes down on the master server (Solaris)!

I had to restore the original file and the db come up again.

Resizing of the unallacated space was done without any problem.

After the upgrade, all my vmware backups failed.

Seems that the privilege

Global -> log event is now necesary

Virtual machine -> configuration -> Disk change tracking is also necesary

since it was still not working I ended up using the account that I also use for the restores with the privileges as described in tech130493. 

But it would be good to know what privileges are only used for making backups. 

Peter

[D_Flood]

Peter,

So there isn't even a 32 bit Java Console for Windows?  That's a showstopper for us then.  Our standard desktop workstations are Win 7 32 bit and there's no way I'm going to switch from auth.conf to NBAC to limit my Managers who "just have to have access!"...

[Peter_Jakobs]

Yep, no more Java Console for Windows 32 bit.

Neither can you change to nbac. nbac uses the windows administration console, and also this one is only for 64 bit.

I understand Symantec that they dropped 32 bit server version, but the Java Console...

[D_Flood]

That is a fatal mistake.  This locks our organization at 7.5 for the forceeable future.

[Peter_Jakobs]

If you need to give some access to some people "just to have access" you can give them access to your Opscenter server (with some very limited privs of course).

But all your admins should change to a 64 bit OS.

[D_Flood]

Sorry Peter, but the real world just doesn't work that way.

99% of the applications that I have to use on an hourly basis, other than NetBackup, work just fine (and in some cases require) 32 bit Windows.  And we just managed to finish rolling out Win 7 x32 on all desktops as of December.  I'd have to add a second machine to my desktop just to administer one application.  There's no way I can justify the cost of a second Windows license just for one application (and no way I'd get permission to run 64 bit Linux on the network at all).

OpsCenter does not provide the information that some of our Senior Management (including the SOX Enforcer) require.  And I don't want or need those persons logging onto the servers directly.  Not providing that information to them means that NetBackup would probably become the less perferred backup option in the company...

So I guess we're stuck with 7.5.0.7 and whatever warts develop with it...unless this turns into the straw that causes Senior Management to start shopping elseware...

[ayhan_sevimli]

Hi all;

One of our customers has a 5220 Appliance that version was 2.5.3. I've started upgrading to 2.6 version. It has 64 TB dedup pool that is used space 36TB. Dedup Conversion step is still going on for 7 hours , process was completed % 45. I think so it'll take 16 hours. 

Regards.

-as

[Peter_Jakobs]

Sorry D, I didn't want to force you to switch to anything, I was just trying to offer a solution in case you need the new functionality of 7.6.

Maybe if enough customers really want/need the Java console on 32 bit that Symantec will do something...

[D_Flood]

No bad feelings Peter.  And aparently some of the frustration I'm feeling at my Managment over another NetBackup related issue bled over to my response.

I guess the next thing for someone to try is see if a x32 7.5 Java will talk to a x64 7.6 server...I know, I know, not supported...but we do a lot of "not supported" in the real world...

[Stanleyj]

Does any one know if there is an upgrade log that I can look at to determine why mine keeps rolling back in the middle? After letting the drives rebuild I attempted the upgrade again and it appeared to be going fine and then about an hour 1/2 into it I noticed it said it was rolling back. This is happening on two different 5200 appliances for me. Now I'm having to wait once again for the drives to rebuild before I can attempt the upgrade to see what the issue is this time around.

[EnriquePereira]

How many time did the install program told that it will take?

With mine, it should be 1,5 hours and has taken almost 5.

[Stanleyj]

Nevermind I think I found them. /log/patch_log_2.6.0.1_ and /log/patch_log_2.6.0.1__ROLLBACK in the event of a rollback.

[EnriquePereira]

Yes, you can check:

[2014-01-07 09:16:52]   /log/patch_log_2.6.0.1_<date> and
[2014-01-07 09:16:52]   /log/patch_log_2.6.0.1_<date>_ROLLBACK in the event of a rollback.
 

This week I had problems upgrading and see more information inside this files.

You can allways open a case to ask for support.