Unsolved
This post is more than 5 years old
7 Posts
0
956
January 10th, 2006 09:00
Relocating Networker from AIX to W2K3
We are planning to move our Networker server from AIX to W2K3. In the process we'll also be changing levels (7.0 to 7.2) and be using a new hostname.
Following the recommendations I've read so far, it would appear I would follow the Disaster Recovery procedures. But I would like to hear from others who have been through such a migration and listen to all the "gotchas" and things to look for.
Thanks!
Following the recommendations I've read so far, it would appear I would follow the Disaster Recovery procedures. But I would like to hear from others who have been through such a migration and listen to all the "gotchas" and things to look for.
Thanks!
0 events found
No Events found!


ble1
6 Operator
•
14.4K Posts
•
56.2K Points
0
January 11th, 2006 00:00
However it can be done - at least it was possible in some instances (like Solaris to Linux I believe or something like that). The best thing here is to get an answer from someone who actually did try this with AIX and win2k3. There are some steps how this could be done via uasm and such steps have been used by 3rd party support providers.
You will face two "problems":
- OS based
- NetWorker based
First one is endian issue (just google big little endian terms and you will get loads of information).
Second one is related to NetWorker and things to address are:
- any index path that was set explicitly
- license migration (hostid will change - this is done free of charge)
Probably something else, but I forgot now.. One thing to recommend is to upgrade old system to 7.2.1 before migrating. Thus you will migrate 7.2.1 -> 7.2.1 avoiding that when actual migration happens problem that you will face is not caused by 7.0->7.2.x migration.
My suggestion would be to test it yourself first. Perhaps you can also keep old server alive just for restores until retention expires (if not too big). Loads of possible options... I believe something like that was already covered in another topic on this forum.
DavidHampson
2 Intern
•
1.1K Posts
0
January 11th, 2006 02:00
The /nsr/res/nsrdb directory is a series of text files so you should be able to build your new server, stop the Networker services, ftp the nsr/res/nsrdb directory from the old server then restart the services. I would suggest you remove any security restrictions from the old Networker config before doing this then reapply them afterwards.
There is an undocumented procedure for migrating the media database from one server to another:
1) Copy the media database to a file using nsrmmdbasm on your old AIX box:
/usr/lib/nsr/nsrmmdbasm -s /nsr/mm/mmvolume6 > savemdb
2) FTP the media database from your AIX box to your new Windows 2003 system (make sure you are performing a binary transfer).
3) Restore the media database to your Windows 2003 NetWorker server:
"nsrmmdbasm -r -2 < savemdb"
NOTE: the "-2" flag is undocumented and relates to the version of the media database.
This flag is appropriate for NW 6.0 through at least NW 7.0 media databases (and I would thus assume to 7.2.x but possibly not 7.3.
You should then be able to rebuild the indexes using nsrck -L6.
I would suggest testing this procedure first and ideally getting your EMC partner in to carry out this work! I would also verify that you are not going to see any block size issues going from a UNIX to a Windows system.
mtmade
7 Posts
0
January 16th, 2006 12:00
Message was edited by: jcross2
ble1
6 Operator
•
14.4K Posts
•
56.2K Points
0
January 16th, 2006 14:00
Article exists - it is here:
http://www.legato.com/resources/bulletins/121.html
Perhaps there has been silent change in mm db XDR format (eXternal Data Representation - the format used by NetWorker to transfer data). See manual for nsr_data for more details. I didn't check manual for 7.3 in that respect, but as far as I know no change was done yet to address this. Keep us informed as this is rather interesting topic.
ble1
6 Operator
•
14.4K Posts
•
56.2K Points
0
January 17th, 2006 02:00
But the same engineering had also couple of problems with same approach (thus making it risky - I have my own theory which says that risky things should not be supported if something is designed to work the way one expects). I would assume steps how to do it did leak from the same source as around year or even a little bit more same has been revealed by ex LGTO engineer on one public mailing list.
The fact is that 2 or 3 years ago such migrations were no - even with proffesional services involved. Then some 12-24 months ago things started to change and uasm approach was mentioned (even no one came forward to say if it works as it was still officially unsupported - as it is now). This is first time I hear that support said to someone this is supported (we are not talking here about us talking to each other or some engineer, but rather official support).
The problem still remains, as you said, that this is risky. If risky it means it could fail - if it fails it means it doesn't work as designed. If it fails again does this mean that is unsupported then? If unsupported how come it was supported before? You can see what I mean with this questions and how difficult this "risky" factor makes whole thing. If support now can say that this is supported then risky factor is gone I would assume - the fact that it needs to be done by proffesional service is ok as normal user should not even think to play with that.
DavidHampson
2 Intern
•
1.1K Posts
0
January 17th, 2006 02:00
The process for cross platform migration was given to me by an EMC engineer about a year back so this was possible with versions earlier than 7.3 - if I remember rightly EMC do not make this process readily available due (1) to it being slightly risky and (2) to ensure they keep consultancy revenue for it.
nsrmmdbasm does not back up the media database files as files which would cause issues if we were restoring to different file systems; it backs up the media database entries as records, so when this is restored the XDR stream is converted into a record and written into the media database, which is why this works (I believe!). It is for the same reason that you can recover your media database even after it has became corrupt, as it is the data that is recovered rather than the files.
David