MDS R80.20 GAIA restore fails

[Kaspars_Zibarts]

Had a first R80.20 backup/restore attempt in the lab and it failed miserably

MDS showed no CMAs although it restore succeeded according to logs. And it was suspiciously fast - approx 10mins that normally took 3-4hrs in R80.10

Digging in restore logs we found following:

Restore script attempts to inflate TGZ file:

Whereas actually it is TAR:

so mds_restore command obviously fails

re-running manually with TAR it succeeds

Anyone else noticed that?

[Kaspars_Zibarts]

Hmmm

mds_backup script in R80.10:

and R80.20:

Will try to modify that and test again 🙂

[Ofer_Barzvi]

Hi all,

I'm taking it offline with Zibarts and will update the thread after completing the investigation.

[genisis__]

I noted this change as well, I think it changed after a Jumbo or specific HF installation.

[Kaspars_Zibarts]

Fixed in $MDSDIR/conf/mds.cpbak, just change the unpacking extension to tar

yes | ./mds_restore *.mdsbk.tgz

to:
yes | ./mds_restore *.mdsbk.tar

Thanks to  @Ofer_Barzvi for speedy resolution!

But most importantly! It's been truly amazing long journey since we upgraded to R80 in early 2017 and noticed MDS backup file increasing quite noticeably. Nobody believed me at the beginning that there is a problem and my cases were promptly shut down. Our backup grew from 3 to 22GB and restore time increased from half an hour to nearly 4hrs..

Now, more than 2 years and many posts and cases later, backup is reduced to 8GB and restore time was 45mins in the lab ESX that's under-powered compare to production one.

Kudos also to @Tal_Kagan for helping over very long time period to sort it out!

And it would not have happened if it was not for the community  - just because of my nagging here case was eventually escalated to the right gang (via Dorit and our Swedish user group) 🙂 

So never give up hope haha

[genisis__]

This should really be in a SK.