Overclock.net › Forums › Specialty Builds › Servers › SANs & VMs: A Question
New Posts  All Forums:Forum Nav:

SANs & VMs: A Question - Page 4

post #31 of 32
I would assume the load balancer checks for heartbeats from the systems (Keepalived is a a linux package which can do this), and disables the routing to the dead host until it's up again.

Or perhaps you can configure a failover SAN in your servers, just like you can configure two DNS servers.
    
CPUMotherboardGraphicsRAM
Q6600 SLACR @ 3.6 GHz Asus P5E Deluxe MSI 6950 2 GB + 9800GT (PhysX) 4 GB White Lake DDR2-800 
Hard DriveOptical DriveOSMonitor
Hitachi 500 GB Sata iHas 120 Windows 7 Pro x64 u2711 (27", 2560x1440, H-IPS) 
KeyboardPowerCaseMouse
Generic Dell Combat Power 750W Aerotech PGS Bx-500 Logitech Rx300 
Mouse Pad
Desk 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
Q6600 SLACR @ 3.6 GHz Asus P5E Deluxe MSI 6950 2 GB + 9800GT (PhysX) 4 GB White Lake DDR2-800 
Hard DriveOptical DriveOSMonitor
Hitachi 500 GB Sata iHas 120 Windows 7 Pro x64 u2711 (27", 2560x1440, H-IPS) 
KeyboardPowerCaseMouse
Generic Dell Combat Power 750W Aerotech PGS Bx-500 Logitech Rx300 
Mouse Pad
Desk 
  hide details  
Reply
post #32 of 32
Thread Starter 
Quote:
Originally Posted by citruspers View Post
I would assume the load balancer checks for heartbeats from the systems (Keepalived is a a linux package which can do this), and disables the routing to the dead host until it's up again.

Or perhaps you can configure a failover SAN in your servers, just like you can configure two DNS servers.
After a little more digging, it looks like the main node's IP can be migrated to the failover node (in a two-box configuration). Very useful, since it means leaving the front filer none the wiser in the event of a SAN node failure.

However, this doesn't help load balancing, since with load balancing you need a pool of IP addresses to choose from, and it seems pointless to assign a separate IP address to the failover node, when it could end up with the primary node's IP address anyway.

To my mind then, it would appear that an effective fault-tolerant SAN cluster will look much like a RAID 5 array, where there are a minimum of three nodes, and at least one node will be a "hot spare" - not part of the active cluster (i.e. no assigned IP address), but ready to be called into action in the event of failure.

Would this be a fair assessment?


Edit:

Just realised that if the spare node didn't have an IP address, the heartbeat system wouldn't be able to locate it in order to fail over to it. duh... Apparently a node can use standard IP aliasing to assume the ID of the failed node. Well, that answers the load balancing question.
Edited by parityboy - 3/8/11 at 1:10pm
Ryzen
(12 items)
 
  
CPUMotherboardGraphicsRAM
Ryzen 7 1700 Gigabyte GA-AB350M Gaming 3 Palit GT-430 Corsair Vengeance LPX CMK16GX4M2B3000C15 
Hard DriveCoolingOSMonitor
Samsung 850 EVO AMD Wraith Spire Linux Mint 18.x Dell UltraSharp U2414H 
KeyboardPowerCaseMouse
Apple Basic Keyboard Thermaltake ToughPower 850W Lian-Li PC-A04B Logitech Trackman Wheel 
  hide details  
Reply
Ryzen
(12 items)
 
  
CPUMotherboardGraphicsRAM
Ryzen 7 1700 Gigabyte GA-AB350M Gaming 3 Palit GT-430 Corsair Vengeance LPX CMK16GX4M2B3000C15 
Hard DriveCoolingOSMonitor
Samsung 850 EVO AMD Wraith Spire Linux Mint 18.x Dell UltraSharp U2414H 
KeyboardPowerCaseMouse
Apple Basic Keyboard Thermaltake ToughPower 850W Lian-Li PC-A04B Logitech Trackman Wheel 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Servers
Overclock.net › Forums › Specialty Builds › Servers › SANs & VMs: A Question