Force a data tiering calculation during a migration

  • 28 May 2019
  • 1 reply

Userlevel 1
Badge +1
We are preparing to migrate about 350TB of data from our existing 3-tier architecture to a Nutanix hybrid solution. Our storage pool is 480TB in size with about 25% SSD before storage efficiency kicks in. We have post-compression turned on as well as cache deduplication. We are going to attempt to migrate 50-100 VM's per night, and we are afraid that we will overrun our SSD tier eventually. Is there a way for me to run a data tiering job, or force the data to destage lower, daily after the migrations are done so that it doesn't impact our future migrations?

Thanks in advance

Best answer by rhunt 30 May 2019, 00:43

View original

This topic has been closed for comments

1 reply

Userlevel 2
Badge +4
Let me start by answering your question and go into why you probably don't want to do it this way....

There is a gflag that can be set to lower the percentage of space used on each of the nodes ssd tier(opslog) this would effectively force move ILM to move the data down a tier.... This is utilization based not time based so as you onboard workloads you are going to inherently have more data moving up and down tiers constantly even if it may need to stay in the opslog because it may be truly hot data... Also only do this at the instruction of an SRE.

From experience of migrating 1000s of servers over to Nutanix on hybrid clusters from 3 tier, I can honestly say you shouldn't worry too much about it. Your performance of the migrations are going to be bottlenecked at your network layer and possibly your migration method.

My advice is let ILM handle the normal destage process from cache to opslog and down to extent as it needs to. Most of the time Curator is set to move data out of Opslog when the average usage is 70% of the total cluster SSD pool.