Forums

Home / Forums

You need to log in to create posts and topics. Login · Register

adding interfaces

Pages: 1 2

Hi,

in my deployment, i have 4 NIC on each node (2*1GB and 2*10 Gb). during the installation wizard i added one of the 1GBs as management (172.19.0.0) and wanted to use one of my 10GBs as backend (on 192.168.11.0 )and the other one as iSCSI (on 192.168.12.0). Now i can reach all nodes on management interface and the backend interface. But when i check my interfaces config, the other 10GB interface is not even assigned an ip. I've tried the regular way of manual IP assignment in Linux(through /etc/network/interfaces and restarting the service and the whole node) but it doesn't work!

after each reboot the new ip assignment does not appear in the interfaces file.

Questions:

  • Am i making any mistakes in my deployment?
  • Is there any other file for network configuration that i'm missing?

You need to update the following files:

/opt/petasan/config/cluster_info.json
/opt/petasan/config/node_info.json

They are straightforward and self explanatory

Quote from admin on January 29, 2018, 4:12 pm

You need to update the following files:

/opt/petasan/config/cluster_info.json
/opt/petasan/config/node_info.json

They are straightforward and self explanatory

 

Thanks.

so would this be the right config:

in cluster_info.json:

{
"backend_1_base_ip": "192.168.11.0",
"backend_1_eth_name": "eth3",
"backend_1_mask": "255.255.255.0",
"backend_2_base_ip": "192.168.13.0",
"backend_2_eth_name": "eth3",
"backend_2_mask": "255.255.255.0",
"iscsi_1_base_ip": "192.168.12.0",                 This line was added
"iscsi_1_eth_name": "eth2",                            This line was added
"iscsi_1_mask": "255.255.255.0",                   This line was added
"iscsi_2_base_ip": "192.168.12.0",                 This line was added
"iscsi_2_eth_name": "eth2",                            This line was added
"iscsi_2_mask": "255.255.255.0",                   This line was added
"bonds": [],
"eth_count": 4,
"jumbo_frames": [
"eth2",
"eth3"
],
"management_eth_name": "eth1",
"management_nodes": [
{
"backend_1_ip": "192.168.11.11",
"backend_2_ip": "192.168.13.11",
"iscsi_1_base_ip": "192.168.12.11",                  This line was added
"iscsi_2_base_ip": "192.168.12.11",                 This line was added
"is_iscsi": true,
"is_management": true,
"is_storage": true,
"management_ip": "172.19.0.44",
"name": "Nodemgmt1"
},
{
"backend_1_ip": "192.168.11.12",
"backend_2_ip": "192.168.13.12",
"iscsi_1_base_ip": "192.168.12.12",                   This line was added
"iscsi_2_base_ip": "192.168.12.12",                   This line was added
"is_iscsi": true,
"is_management": true,
"is_storage": true,
"management_ip": "172.19.0.45",
"name": "Nodemgmt2"
}
],
"name": "Peta1"

 

and in node_info.json:

{
"backend_1_ip": "192.168.11.11",
"backend_2_ip": "192.168.13.11",
"iscsi_1_ip": "192.168.12.11",
"iscsi_2_ip": "192.168.12.11",
"is_iscsi": true,
"is_management": true,
"is_storage": true,
"management_ip": "172.19.0.44",
"name": "Nodemgmt1"

 

And the same story for other two nodes (of course with different IPs)

Please clarify why iscsi2 exists? Could I don't use it at all? Or it is used just because to be able to mount some target on the separate subnet?

It is used with mutipath io ( mpio ), it is the standard way to provide high availability to iSCSI clients.

If you have a highly available storage systems but the port/link/switch between the client and the storage systems goes down, the client io will stop even if the storage system is highly available. bu having a separate network path from client to system it will create redundancy for the network link.

You do not need to use it in PetaSAN if you do not need this availability. just define both subnets on same nic and create you iSCSI disks on 1 subnet.

One more question regarding small home setup: can I combine two backends into one? How much data is on that "IO iSCSI Target layer to communicate with the Ceph storage engine."?

Yes you can combine many subnets into a single interface. the deployment wizard allows you this.

For a description of subnets and their traffic please see:

http://www.petasan.org/forums/?view=thread&id=158

Hi,

I have only 2 NICs in each node. One NIC is 1 Gbps NIC and the second one is 10 Gbps NIC.   I do not have provision to add additional NICs in these nodes.  Can I use these nodes for setting up PetaSan?  If so what is your recommendation of assigning these network interfaces to different subnets?

Making decision is a real quick.  Quick response is highly appreciated.  Thank you.

 

You can put your management on 1G and all the others on 10 G

Thank you for your quick response.  Currently all my client systems are in single subnet connected under 1 G network.  A single 10 G network switch is used to connect to 10 G NICs of nodes.  In this case do I need to configure all the 5 subnets.  If so will the following be ok.

  • iSCSI-1 - 1 G
  • iSCSI-2 - 1 G
    • Do I really need two iSCSI nodes.  We have single subnet in our network currently (192.168.0.x).
  • Backend 1 - 10 G
  • Backend 2 - 10 G
    • Here, as well do I really to create two Backend subnets.  Backend subnet currently is (10.1.1.x)
  • Management - 1 G
    • I believe that this need to be in the same subnet of iSCSI (192.168.0.x).  Is that right?

Please advise. Thank you.

Pages: 1 2