Unsolved
This post is more than 5 years old
3 Posts
0
462
Node replacement vs migration
We have a 24 node cluster (2 cubes). The 1st cube is the first 16 nodes. All nodes are gen 4LP. four of the nodes in the 1st cube are 3TB models which will be EOSL 2/2014. The rest are 4TB or 8TB models.
So the question is...
Can an in place replacement of those 4 nodes be done without having to migrate the whole cube? ie replace them as if they had failed with newer models and let it rebuild the data on them?
Anonymous
5 Practitioner
5 Practitioner
•
274.2K Posts
0
October 20th, 2013 11:00
C* 4.1 code supporting migration of individual nodes. First migrate first 4 nodes to rest of the cube. When the migration finishes, migration script automatically removes nodes from the cluster. Uninstall old nodes from the rack and install and power up new nodes. They will automatically join the cube as c001n01-04. That's all.
For a quick tip, if you don't have enough space on the cube, you can migrate 2 of them first for example, install just two new nodes, then migrate other two.
Please feel free to ask, as I just migrated 6 nodes in a cube. It was a scenario like yours.
Best regards,
Cenk
emcmobile_ws
19 Posts
0
October 31st, 2013 04:00
testing please ignore