Artesyn Centellis 4100 Installation and Use (2015) - Manual (Page 192)

Configuring and Operating the System
Centellis 4100 Installation and Use (6806800D82E) 194
6.7 Redundancy
The following sections outline the interaction between redundant shelf managers and
redundant hub blades. This interaction is not described within the AdvancedTCA Specification
PICMG 3.0. The following sections also describe redundancy scenarios with redundant shelf
managers and system managers.
6.7.1 Cold Standby
In Centellis 4100 systems the shelf managers work in redundancy mode, this means that the
HPI daemon runs on the active shelf manager and is started on the former stand-by shelf
manager once it has become the active one.
When a former standby shelf manager becomes active the following steps are carried out:
 Removal of the ShMC IPMI address from the former active shelf manager
 Assignment of the ShMC IPMI address to the local IPMC
 Assignment of the active virtual IP address
 Sending a gratuitous address resolution protocol (ARP)
(broadcast to update arp caches)
 Starting the HPI daemon which performs HPI resource discovery for all resources of the
AdvancedTCA shelf
 Storing of the resource IDs which were assigned to the former active shelf manager. Thus,
it is possible to reassign these IDs when the session is reestablished.
 Closing of all open HPI sessions on the client's side. They have to be reestablished. Any
blocking HPI calls have to be avoided (for example, saHpiGetEvent) and calls with timeouts
have to be used instead.
6.7.1.1 Heartbeat
The shelf managers use two private interfaces to heartbeat each other. The heartbeat interval
is configured to 400 ms and the partner is considered inactive when no responses are received
after five retries within 2 seconds.