VDP 6.1 STORAGE CHALLENGE

[chrisazzi]

I have a vpd storage domain showing clients consuming terrabytes of data.... these are clients which i have deleted using the web interface and i have also deleted all restore points.

however they are all still consuming space on my storage destination.

I have run capacity.sh and it shows the top 3 clients.... i need instrutions please on how to go through and delete these foreign clients.

thank you.

[chrisazzi · Accepted solution]

great community. hahaha

[SavkoorSuhas]

So usually we have something called as data stripes in your data0? partitions

In these stripes you have your data chunks.
Each stripe is roughly about 250MB.

Now, when you backup a client, the data is striped and stored across these stripes under all your available data0? partitions.

Similarly when you delete a restore point for VM, the space is not immediately reclaimed.

Two processes take place (They are kind of interrelated)

The first (kind of) process is garbage collection which does the "space reclamation"
In the back end this does not actually reclaim space. What this does is, it goes through all the data stripes and checks if the stripe was modified.

Modified being your restore point removed from GUI or CLI

If the stripe was modified by deletion then it is "marked" for reclamation.

Async Crunching is the one that does the actual space reclamation.

Consider Async Crunching something like this.

You have a bus completely filled and new set of people are trying to get in but unable to.

Then, right in the middle of the bus, a bunch of people vanish. Yet, the people at the door are unable to get in unless the space in the middle is reclaimed.

This is what async crunching does.

So all in short, these tasks occur in the maintenance window of the VDP. So you will ideally need to wait for a day or two for space reclamation to complete.

"or two"? Well, because GC has a switch called as "pass"

Pass is nothing but how many times the GC has to go through each stripe to make sure nothing else is left to be "marked"

So, if a large set of data is deleted, then GC might need to perform multiple passes. Keeping in mind GC runs only 1/3 of first maintenance window as well, you have IC and HFS to be executed as well.

To see how much is reclaimed run

status.dpn

This will give you an overview of amount space reclaimed by GC

To see if GC is running, behold another tool!

dumpmaintlogs --types=gc --days=n | grep -i "<4"

where n is an integer (of course!)

If GC is not doing the job as expected, then well VMware Support is where you need to go for!

Suhas