#Exapanding StorageGrid and change EC policy
1 messages · Page 1 of 1 (latest)
Here is link to NetApp docs that can help you et started
few steps to consider
when adding new storage nodes
i assume you currently have 3 ADC's or do you have 4?
when adding new storage nodes, make sure you are adding storage nodes with same drive types as your existing storage nodes
if for some reason you are planning to add storage nodes with drives with higher capacity than your current nodes, then additional planning is required to prevent any issues due to unbalanced usage of storage nodes
when expanding existing grid, it is always a good practice to add all your new nodes as ADC's .
it just makes your future planning to expand a lot easier and process of decommissioning a lot more seamless
As for EC planning, please plan for EC profile that is one less than total number of storage nodes and for effective optimization of usage, make sure to choose the option to apply new EC to all objects from day 1
This would have negligible impact on performance.
4+! gives you 25% overhead and 6+1 further reduces overhead to 16%
hope this helps
Those are very helpful messages. It sounds you know SG, I have following follow-up’s:
- What number of the recommended storage nodes we should upgrade to based on what we have below:
4 storage nodes and 3 ADC’s(not sure of why 3 not 4). Each node has 5 draws, 60 drives(2 x 744.71GB SSD’s; 58 x 3720.52GB’s). - With EC=6+1, we can have less overhead, and therefore to be able to storage more data, that is what we need. Are there concerns then about lower tolerance on disk failures?
- Based on reading, the way to change EC policy (EC rebalancing) is to run commands to run on Admin node, do you remember what they are?
what model of NetApp SG do you have, i mean what is teh model of teh 4 storage nodes
SG6060 / E2800
3 ADC's per site/grid is a requirement which is why it is 3 but it not a limitation
This link will provide info on recommended number of storage nodes and find the sweet spot per your requirement
i assume this is a single site configuration
Yes, single site
we need to think about striking a balance between fault tolerance, % of overhead we want to achieve
and your budget
option 1- add 3 more nodes, create 4+! EC profile, add the new nodes with ADC profile
this will reduce your overhead to 25%
and give you options to expand at will by adding storage shelf's as opposed to adding new storage nodes
if you want lower overhead, like 16-17%
Adding new storage nodes: storage nodes should be of the same or higher configuration. In this case you need to choose SG6060 or higher when you add additional storage nodes
for 16-17% overhead, best practice: Add 4 more storage nodes, create EC profile of 6+1
if budget constraints, then add 3 nodes instead and create 6+1 EC profile
in all cases makes sure to choose the new storage nodes for the role of ADC because you can only choose the role of a node when installing it for first time
rebalance-data start --site "site-name"
is the command
I am trying to understand options you have provided for me:
option 1 - add 3 more nodes, create 4+1 EC profile, overhead --> 25%
option 2 - add 4 (preferred) or 3 nodes, 6+1 EC, overhead-->16-17%.
in all options, nodes should be SG6060 or higher, new nodes should have the role of ADC.
Am I understanding you correct?
What did you say about adding storage shelves as opposed to adding new nodes? Is this option 3?
Thanks for your time!
Disregard my message about adding expansion shelf's. Summary of options below
option1
Add 3 additional nodes, that makes your grid to have 7 total storage nodes, new nodes to be added with ADC role
If you are concerned about impact to disk failures, then 4+1 EC rule can be implemented with a trade off of 25% overhead
If you are not so concerned about disk failures, then 6+1 EC Rule can be implemented to achieve lowest overhead of 16-17%
Option 2
Add 4 additional storage nodes, that makes your grid to have total of 8 nodes, new nodes to be added with ADC role
implement 6+1 EC rule and achieve lowest overhead of 16-17%
Option 2 would be the recommended configuration to achieve the lowest possible overhead and comply with all NetApp best practices
In all options the new nodes should be SG6060 or higher.
Hi, did you have any additional questions on the summary i provided
feel free to connect @ harold@lcubeddataservices.com for any additional consult
I am good for now. Your message are very helpful. Thank you!
ok