What is the best backup strategy for Windchill?

dmohan · ‎Jan 19, 2016

I know that there are some other discussions on this same topic, but I had a few questions for which I could not find answers.

Our Windchill vault is around 1.7 TB in size with master and 9 replica servers scattered around the globe.

Our backup procedure is as follows for the master server.

Full backup of the server once a month which includes all files systems and the vaults.
a partial backup of the masteruploadfolder (cache vault) and mastervault every Friday (which is around 900 GB)
Daily Differential backup (which is around 1-10 GB)

Since the partial backup is around 900GB, it has started choking our network when taking a backup and extends into our regular working hours, causing problems to other business applications.

Once a month we remove the unreferenced files but that does not improve the situation much.

We have set the property wt.fv.fvFileThreshold=50000. So our current mastervault_Folder sequence is 110 and that of the cache vault is 15. All the previous folders from 1 to 109 on master and 1 to 14 on cache vault are set to Read only by Windchill.

I have noticed that even though the the folders are read only, files are being removed from these locations when we run the utility to remove the unreferenced file.

So here are my questions.

Is it OK to just backup all the folders in the vault once and then just backup the writable folders from the vault every week (cache vault and master vault)? If this is the case we can reduce the backup size from 900 GB to ~100 GB
If we just backup the writable folders, there will be a lot of unreferenced file in the read only folders that was backed up. Is it OK to just leave it as it is in the backup? Will it create any problems in the future if in case we need to restore the vaults from these backups?
Other than the mastervault folders and the cache vault folders (the ones that are writable) is there any other locations that needs to be backed up daily?

Hope to hear from the experts out there.

rkandepu · ‎Jan 20, 2016

Hi Dinil,

You may do like this:

Backup Plan:

1. Take a full back of vaults (All folders including cache&master vaults) and make it as 'Base Point'

2. Then take differential backup on daily basis

3. You can avoid partial back-up of masteruploadfolder that happening on every Friday.

Note: Follow the same approach for all replica & master servers.

Restore Plan:

At any point of time, if you need to restore you can use both Basepoint & respective differential backups to restore it to specific date - based on your input the whole referential data for one month will consume around 300GB.

Best Regards,

Ravi Kandepu.

BineshKumar1 · ‎Jan 20, 2016

Hello Dinil,

I agree with Ravi. You don't have to take a periodic backup of read-only folders, after you have taken a one time backup of current read -only folders, keep them as base point and exclude those folders from daily differential backup schedule. You can update the exclusion list based based on the folder status.

When you run remove un-referenced, select move option and gather the file name of unreferenced files from the all .unreferenced directories and delete all those files from the backup (provided you don't need any of those unreferenced files in future). Most of this tasks could be scripted - if you are using a linux system, this is very easy will shell scripts. If your system responds well to Windu tasks, then you can also use un referenced file report in WinDU

As for the backup choking the network, try creating dedicated network adapters for backup and use backup LAN to route this traffic. If you are using SAN, there are many ways to optimizing the backup (like compression and so on) without affecting the normal network.

One location which I would strongly recommend to backup along with vault would be the Solr Index data directories (even though it is a secondary data repository). This directories grow big over the time and if you ever lose it you will have to re-index the entire date which will be a huge effort on a live system.

Thank you

Binesh Kumar