On 6/29/20 11:20 AM, Gordon Messmer wrote:
...
In the event of a power loss, many servers will boot faster than the
managed Ethernet switch they are attached to. Systems managed by
network-scripts may not set up their network because there is no
carrier at the time that networks-scripts start up.
Network-manager, on the other hand, will set up networking whenever
the interface becomes ready.
At $dayjob we got to see the advantages of this up close and personal a
few days ago when we had a hard failure in the voltage regulator/exciter
loop of one of our three primary generators during a utility power
failure. First failure of this kind since that datacenter was placed
online in 2008, and only the second power out event since 2008. (While
critical systems have dual generator backup, and all systems have UPS
with a few minutes run time, the systems affected by this weren't
previously classed as 'critical' enough to have dual feeds, although
that might change in the next few weeks). So, we have a virtualization
host in an IBM Bladecenter with iSCSI shared storage on some EMC
Clariion LUNs through a Cisco 7609 switch/router with RSP720. Anyone
who has ever dealt with these items in a cold boot situation probably
knows how the litany goes at this point.....
Prior to the NetworkManager intelligence, I would have to manually reset
the host in the Bladecenter after the EMC was up and after the 7609 was
up. Not any more. Nice. (Clariion can take 15-20 minutes to come all
the way back up on a hard power fail; 7609 can take 10 minutes if not
longer. Bladecenter takes 5 at most.)
_______________________________________________
CentOS mailing list
CentOS@xxxxxxxxxx
https://lists.centos.org/mailman/listinfo/centos
