HPE 1950 static routing and failover

[Neodym]

Hi! We are currently planning a switch transition from Cisco to HPE. For our servers and user workstations we have decided for the 1850-48 4XGT (JL171A). We need two agreggation switches for the (5) 1850s that can do static routing (as subnet gateways) and failover. We are considering two 1950 12XGT 4SFP+ (JH295A) configured as a stack (IRF in HP lingo?).

The question is, if one stack member fails, does the standby member continue routing using the configured VLAN IP interfaces? Also, how fast is the transition? Are we going to experience any downtime? Can we perform maintenance (reboot or firmware updates) on the stack without disrupting services?

Thank you in advance!

[parnassus]

Hi, most of your questions have answer on this thread...clearly if you want to benefits of the resiliency provided by IRF (on HPE OfficeConnect 1950 Switch series HPE masked the Comware IRF calling it "stacking") you should consider to have all downlinks to access switches formed with LACP based Link Aggregations Groups, known as LAGs (Port Trunking in HP/HPE jargon), where each LAG has one link (or more links) to IRF Member 1 and the other link (or links) to IRF Member 2.

With regards to network disruption to consider during maintenance operations (software updates)...if LAGs are deployed between IRF Members and Access Switches (or Servers, if those are directly connected to IRF Members)...the disruption should be minimized because each Switch/Server will continue its operations through the LAG's link active to the active IRF Member while the other is updating.

[Neodym]

Thanks for the answer! I was planning to connect the switches like you descibed in order to provide redundancy and high availability.

Regarding failover or a planned maintenance, I would like to avoid experiencing the same problems we have with our current Cisco SG500s. They also advertise stacking (which works), but on the high availability department they fail miserably. When one of the 2 members goes offline (or gracefully switching master), it takes about half a minute before the stack is "reconfigured" (?). During that time all services are down. When asking in the Cisco forums if that behaviour is normal, the answer we got was more or less "what do you expect when you buy so cheap?!". 

So, I would like to avoid buying HPs version of "SG500", if you know what I mean. I am looking for milliseconds of downtime when performing maintenance or in case of switch failure.