The paper above assumes NFSv3, which is a stateless protocol.
For NFSv4, which is a stateful protocol, each NFSv4 client is responsible for its own recovery should the server go down, or the server's IP repoints to another node.
I hope this provides you some insight. If not, please elaborate on the question.
I have the same question. It is common knowledge that it is good practice to create 2 pools. one for static and one for dynamic. In a traditional setup CIFS [stateful] (SMB1, SMB2) would utilize the static pool and NFS [stateless] (nfs v3) would use the dynamic pool. So given this context, the question is.
Now that NFS v4 is stateful, should this utilize the static pool or the dynamic pool? What are the pros/cons of this decision.
This would best be broken down on a stateful vs stateless connection. As the stateful connections (SMB and NFSv4) will respond the same way in an instance where a nodes goes down. As outlined in the document RobChang-Isilon mentioned on page 10 under "Client Connection Behavior"
NFSv3 being stateless will seamlessly fail over another node or IP. IE Command will fail but it will try again.
The difference between static and dynamic when it comes to smartconnect zone is the ability to automatically rebalance the IP's to another node. Outlined on Page 10 under "Rebalance policy." A down side to dynamic being pending the way the policy is set an IP can move to another node, causing connection interruptions on stateful clients. See the "Rebalance policy" to see when this can occur based on policy being used.
Static IP assignment would be as follows:
Node 1 with IP 10.10.10.1 goes down that IP will not be redistributed to another node. Meaning if i have a static connection to a share (\\10.10.10.1\ifs) I will not be able to reconnect to the share until this node is back online.
Either way if smartconnect zone name is being used, it will only hand out IP's that are up, so reconnecting/retries will be directed to an IP that is up.
Regardless of dynamic or static configuration the NFSv4 being stateful connection will be broken because the state is lost.
khkris There would be no features lost by using Dynamic compared to Static.
We do also have a best practice document on smartconnect zones and recommended configurations:
RobChang-Isilon
136 Posts
1
March 13th, 2015 15:00
Hi Khkris,
I'm not fully understanding your question. By "specific features", do you mean features within OneFS?
Please see if pages 8 and 9 from this whitepaper answers your question -- http://www.emc.com/collateral/hardware/white-papers/h8316-wp-smartconnect.pdf
The paper above assumes NFSv3, which is a stateless protocol.
For NFSv4, which is a stateful protocol, each NFSv4 client is responsible for its own recovery should the server go down, or the server's IP repoints to another node.
I hope this provides you some insight. If not, please elaborate on the question.
Thanks!
adminguy
1 Rookie
•
17 Posts
0
July 31st, 2015 11:00
I have the same question. It is common knowledge that it is good practice to create 2 pools. one for static and one for dynamic. In a traditional setup CIFS [stateful] (SMB1, SMB2) would utilize the static pool and NFS [stateless] (nfs v3) would use the dynamic pool. So given this context, the question is.
Now that NFS v4 is stateful, should this utilize the static pool or the dynamic pool? What are the pros/cons of this decision.
Stdekart
104 Posts
0
August 6th, 2015 11:00
securityjm,
This would best be broken down on a stateful vs stateless connection. As the stateful connections (SMB and NFSv4) will respond the same way in an instance where a nodes goes down. As outlined in the document RobChang-Isilon mentioned on page 10 under "Client Connection Behavior"
NFSv3 being stateless will seamlessly fail over another node or IP. IE Command will fail but it will try again.
The difference between static and dynamic when it comes to smartconnect zone is the ability to automatically rebalance the IP's to another node. Outlined on Page 10 under "Rebalance policy." A down side to dynamic being pending the way the policy is set an IP can move to another node, causing connection interruptions on stateful clients. See the "Rebalance policy" to see when this can occur based on policy being used.
Static IP assignment would be as follows:
Node 1 with IP 10.10.10.1 goes down that IP will not be redistributed to another node. Meaning if i have a static connection to a share (\\10.10.10.1\ifs) I will not be able to reconnect to the share until this node is back online.
Either way if smartconnect zone name is being used, it will only hand out IP's that are up, so reconnecting/retries will be directed to an IP that is up.
Regardless of dynamic or static configuration the NFSv4 being stateful connection will be broken because the state is lost.
khkris There would be no features lost by using Dynamic compared to Static.
We do also have a best practice document on smartconnect zones and recommended configurations:
https://support.emc.com/docu58740_Isilon-External-Network-Connectivity-Guide---Routing,-Network-Topologies,-and-Best-Pra…
Page 21 goes over "Protocols and Smartconnect zone types" Which smartconnect zone type is recommended for a specific protocol.