The downside of running low of free space

  • 3 October 2015
  • 2 replies
  • 798 views

Badge +7
This isn't really installation and configuration but it was related to my protection domain snapshot issue. So, when I ran low on free space, the metadata increased becasue there were still tons of blocks out there to keep track of. So, this combination of low space and threshold breaking metasada size exceeds the preconfigured (metadata) size limit before kicking off automation actions.

I don't really understand these failsafe actions fully but both inline compression and deduplication fingerprinting (I use only performance tier dedup but my true love is compression) are disabled. I am not so sure about compression but I can see deduplication fingerprinting adding to metadata. The kicker, to me, is that space conservation via deduplication and compression is more important when I am low on space.

The recovery from these automatic actions is to free space then call support to re-enable the disabled features. I am waiting for an answer why they do not automatically re-enable and more so, I am asking for a "fix" to this behavior. I prefer self-healing, which is fundamental to the architecture, than manual interference.

So, beware of your metadata sizes and keep in mind the aded affects of snapshots. Have your serial number handy to open a support case to get your space savings powers turned back on.

~ Joe

This topic has been closed for comments

2 replies

Userlevel 2
Badge +13
Hi 

This is why Compression is always recommended and dedupe only in specific use cases like full clone moved desktops.

If you are using cloning you get the benifits of 100% dedupe without any metadata overhead.
Badge +7
I also do not use deduplication to reduce the memory overhead for the CVM. CMV memory requirements can be expensive on a 128GB host. I strongly encourage everyone to start with 256GB. Starting with 128GB is a choice I regret. I run about 50 active VMs/host and 60 or 70 total in inventory.

On one of my clusers, an 8 node 3000 series, I have 460 VMs using 23TB. I get 1.63:1 compression saving for approx 9TB raw capacity (or 11.5TB logical saving 4.5TB logical with 10TB logical free).