Delete a node from cluster
ek-media
3 Posts
May 9, 2017, 1:10 pmQuote from ek-media on May 9, 2017, 1:10 pmIs there a way to remove a node from a running cluster ?
Is there a way to remove a node from a running cluster ?
admin
2,930 Posts
May 9, 2017, 2:48 pmQuote from admin on May 9, 2017, 2:48 pmYou cannot remove a running node.
If the node is down, offline or dis-connected from cluster:
If it is not a management node, meaning it is node number 4 or greater, it can be removed. All local disk resources it was handling will also be removed, it is possible that before you delete the node to physically plug some of the OSD disks and place them in other nodes, otherwise Ceph will in the background re-create the data replicas somewhere else.
If it is a management node, you cannot remove it. If a management node fails beyond repair, you can get a brand new machine and specify "Replace Management Node" in step 1 of the deployment Wizard.
It is not possible to logically remove a management node without replacement, Ceph and Consul and many other clustering software should not be left too long running with only 2 nodes, else other failures happen it could lead to split brain conditions.
You cannot remove a running node.
If the node is down, offline or dis-connected from cluster:
If it is not a management node, meaning it is node number 4 or greater, it can be removed. All local disk resources it was handling will also be removed, it is possible that before you delete the node to physically plug some of the OSD disks and place them in other nodes, otherwise Ceph will in the background re-create the data replicas somewhere else.
If it is a management node, you cannot remove it. If a management node fails beyond repair, you can get a brand new machine and specify "Replace Management Node" in step 1 of the deployment Wizard.
It is not possible to logically remove a management node without replacement, Ceph and Consul and many other clustering software should not be left too long running with only 2 nodes, else other failures happen it could lead to split brain conditions.
jnickel
12 Posts
November 19, 2019, 9:36 pmQuote from jnickel on November 19, 2019, 9:36 pmI really like the PetaSAN product and you guys should be proud of what you have accomplished!
This is just a suggestion - why not put information like that above into the Admin manual? Explaining things like that would prevent questions...I really do look in the manual first for information, but often I find I have to look in the forums for the answer.
Another example is the explanation of the differences between Filestore and Bluestore engines.
Another example is the explanation of the differences between Replicated and EC - and what is EC (Erasure Coding, I know, but it is not explained in the manual).
Jim
I really like the PetaSAN product and you guys should be proud of what you have accomplished!
This is just a suggestion - why not put information like that above into the Admin manual? Explaining things like that would prevent questions...I really do look in the manual first for information, but often I find I have to look in the forums for the answer.
Another example is the explanation of the differences between Filestore and Bluestore engines.
Another example is the explanation of the differences between Replicated and EC - and what is EC (Erasure Coding, I know, but it is not explained in the manual).
Jim
Delete a node from cluster
ek-media
3 Posts
Quote from ek-media on May 9, 2017, 1:10 pmIs there a way to remove a node from a running cluster ?
Is there a way to remove a node from a running cluster ?
admin
2,930 Posts
Quote from admin on May 9, 2017, 2:48 pmYou cannot remove a running node.
If the node is down, offline or dis-connected from cluster:
If it is not a management node, meaning it is node number 4 or greater, it can be removed. All local disk resources it was handling will also be removed, it is possible that before you delete the node to physically plug some of the OSD disks and place them in other nodes, otherwise Ceph will in the background re-create the data replicas somewhere else.
If it is a management node, you cannot remove it. If a management node fails beyond repair, you can get a brand new machine and specify "Replace Management Node" in step 1 of the deployment Wizard.
It is not possible to logically remove a management node without replacement, Ceph and Consul and many other clustering software should not be left too long running with only 2 nodes, else other failures happen it could lead to split brain conditions.
You cannot remove a running node.
If the node is down, offline or dis-connected from cluster:
If it is not a management node, meaning it is node number 4 or greater, it can be removed. All local disk resources it was handling will also be removed, it is possible that before you delete the node to physically plug some of the OSD disks and place them in other nodes, otherwise Ceph will in the background re-create the data replicas somewhere else.
If it is a management node, you cannot remove it. If a management node fails beyond repair, you can get a brand new machine and specify "Replace Management Node" in step 1 of the deployment Wizard.
It is not possible to logically remove a management node without replacement, Ceph and Consul and many other clustering software should not be left too long running with only 2 nodes, else other failures happen it could lead to split brain conditions.
jnickel
12 Posts
Quote from jnickel on November 19, 2019, 9:36 pmI really like the PetaSAN product and you guys should be proud of what you have accomplished!
This is just a suggestion - why not put information like that above into the Admin manual? Explaining things like that would prevent questions...I really do look in the manual first for information, but often I find I have to look in the forums for the answer.
Another example is the explanation of the differences between Filestore and Bluestore engines.
Another example is the explanation of the differences between Replicated and EC - and what is EC (Erasure Coding, I know, but it is not explained in the manual).
Jim
I really like the PetaSAN product and you guys should be proud of what you have accomplished!
This is just a suggestion - why not put information like that above into the Admin manual? Explaining things like that would prevent questions...I really do look in the manual first for information, but often I find I have to look in the forums for the answer.
Another example is the explanation of the differences between Filestore and Bluestore engines.
Another example is the explanation of the differences between Replicated and EC - and what is EC (Erasure Coding, I know, but it is not explained in the manual).
Jim