Autobalance - by percentage or by quantity?

Hi,

Thanks for your question. It should be by the number of drives used and the protection policy and not by the amount of data. So yes the first one. https://dell.to/3PNjR64

Let us know if there is anything else we can help you with.

Thanks, but I'm a little unclear as to what you're saying. "By the number of drives and the protection policy" does not sound like it takes into account the size of the drives, just the number (which would be the same for the same model node, even if the drives are bigger). I could not find the relevant info in the link you sent.

So maybe a clearer way to ask the question is does Isilon take into account the size of the drives when it auto-balances, and thus put more data on the nodes with the larger drives?

It depends on if you are using file or full array rebalancing.

It will only use the smallest amount to keep them even. with full array rebalancing. Page 28 https://dell.to/3NNE2yl

The goal of the AutoBalance job is to ensure that each node has the same amount of data
on it so that data is evenly balanced across the cluster. AutoBalance, along with the
Collect job, runs after any cluster group change, unless there are any storage nodes in a
down state.
Upon visiting each file, AutoBalance performs the following two operations:
• File-level rebalancing
• Full-array rebalancing
For file-level rebalancing, AutoBalance evenly spreads data across the cluster’s nodes to
achieve balance within a particular file. With full-array rebalancing, AutoBalance moves
data between nodes to achieve an overall cluster balance within a 5 percent delta across
nodes.

Thank you, but that still doesn't answer my question. "5 per cent delta across nodes" could mean 5% of the total amount of data (e.g. 60TB +-5% per node regardless of node size) or it could mean each node is within 5% of the others in terms of used capacity (i.e. every node has roughly 13% capacity usage, which for the bigger nodes = 80TB actual usage but for the smaller nodes means 40TB actual usage).

Honestly it doesn't make much sense to do the former, since you could then theoretically fill the cluster enough to completely fill the smaller nodes while still having free space on the larger ones, rather than keeping free space available across all the nodes regardless of size.

Which probably answers my question right there, though it would be nice to see it in the documentation/hear it from Dell in clear terms.

You are not supposed to mix drive sizes within a pool. So if you have 1 node of 300GB drives and 3 of 600GB you are most likely only going to have the useable space of 4 with 300GB drives and it won’t use half of the space on three of the nodes. If it lets you create a cluster at all.

Here is more information about how it should work.

 https://dell.to/3O6ysIp

and https://dell.to/3O537Gi

With 4 nodes and 4 nodes you would have separate pools one of 300GB and one of 600GB and they would split the data among the pools.