Will the NetApp nodes performance get degraded if tiering destinatoin SG is getting full?

[dim frost]

We have 8 nodes StorageGrid, and it is so full, 4 of them became read-only. The SG is used for aggreates tiering target.

Will the almost full SG bring down the performance of the aggregates or nodes as the tiering source? If yes, because the NetApp cluster performance is the top priority, I would want to stop writing / putting data to SG, is there anyway to achive that?

Here is what I am thinking, as long as there are available space on any nodes, data will be distributed there via Load Balacer, so, the performance on NetApp cluster shouldn't get degraded, in my opinion.

[clever geyser]

FabricPool relies on Erasure Coding, so you just aren't going to be able to "tier" new data.

[dim frost]

What if all nodes in SG are 100% full, in this scenario, you cannot tier any data at all, would that effect the performance, because a lot of volume are trying to tier but failed, those time on failures could bring down the performance. Correct?

[clever geyser]

i don't know enough about retries between Ontap and SG, but it's just not going to be ideal. I guess your logs will give you some hint about problems already.

if you have some "low load" periods, you might try to increase the "unreclaimed-space-threashold" a bit, but it is going to cause a bit of churn. Tiering is surprisingly CPU intensive on older platforms

you will free up potentially a bit of space and it will sort of "rebalance" things just by the creation of new objects, but YMMV, depending on how much is actually invalid in the objects to start with