Indexing Guest files different size

[prubner]

Hello,

In my demo I have different sizes of the Index files if making backups with Application aware or not.
Indexing with application aware is 10 times smaler than without.
Is it a bug?
Veeam Support say this Veeam Support - Case # 03527739:
- Hello Peter,
- Thank you for your request to Veeam Support. I tested “Indexing” in my virtual lab, and can say that this behavior is “by design”. I have the same results with the indexing files size. (Screenshot 1)
- Maybe, it is connected with VSS Snapshot creation during the AAP process:
- https://helpcenter.veeam.com/docs/backu ... l?ver=95u4
The info in the link above does not say anything about the filesize. I don't think that there are more files changed without application aware so that the index file grows.
Size with application aware: 7.983 KB
Size without application aware: 73.796 KB

Who can explain this to me?
Peter

[foggy]

Hi Peter, just an assumption: when VSS snapshot is created, indexing parses MTF file directly to retrieve data, while without VSS it needs to enumerate all files on the guest OS file system and presumably, requires more data to be stored for each file, which, in case of large amount of files might result in considerable index file size difference.

[prubner]

Thanks für the fast answer.

Do you know something about indexing in new NAS feature of v10?
One of my customer would like to backup Netapp filer and the users shoud make a restore with the Enterprise Manager Self Service Portal.

[foggy]

I had a brief discussion with dev folks and according to them both mechanisms should produce ~similar amount of indexing data, at least it was their intention. The only major difference between them is that the one without VSS is not able to collect file owners and stores information for common files (like AppData, for example) for each user separately. This shouldn't result in that big difference though. We would appreciate if you could continue working with support to identify the reasons of the behavior you're seeing, so we could address it in case it is an unexpected one. Thanks!

I cannot comment re: NAS backup indexing though.

[PTide]

Peter,

I found no logs attached in the case, only .doc file that contains nothing but two pictures. Kindly collect logs and upload those on the ftp that the engineer has given to you.

Thanks!

[prubner]

Indexing bug identified.
The support team found two bugs in the code.

-Hello Peter,

-Thank you for waiting. We and our colleagues from the QA department found out the root cause for different index files size.
-The reason for different file size is two found bugs.
-First of all, for FF/FN algorithm (Indexing without AAP) Veeam index symlinks in the excluded folders despite on the exclusion settings.
-The second bug that affects indexing - folder’s symlinks are indexed along with the original folders. And it’s also for FF/FN indexing mode.
-Thank you for your attention to the indexing file size – it helped us to find new bugs in the indexing mechanism. We really appreciate when our customers help us to improve Veeam products.
-We will fix these bugs in one of the future updates. As “workaround” use AAP for the jobs where you need indexing. In this case, we avoid using FF/FN indexing mode. (Only for NTFS and VMware)

Thanks for help
Peter

[prubner]

How can I calulate the space I must reserve for the indexing files.
I found this in the Best Practice website:
"file after compression is approximately 2 MB per 1,000,000 files for a single VM restore point on the Enterprise Manager server.”
But this does not match with my indexfile size.
Support answer:
In general, value “2MB per 1.000.000 files” – it’s an average value that was obtained in the test environment. And it’s a big difference between test and production environment. So in your environment result for indexing is different - and it's a "by-design" result.
It does not help.
My customer wants do backup 13 Fileserver VM with many TB filespace.
We must calculate the VBRCatalog size for this szenario.
Peter

[foggy]

There is no tool to calculate the size of index data, you can only approximate based on some "typical" VM.