Scheduled core04.tor2 & dist04.tor2 network maintenance - 21-Aug-2020

[porcupine]

Hi Folks,

Please be advised of the following maintenance window:

Maintenance Window: Friday August 21st, 2020 - 11:00pm - Saturday August 22nd, 2020 - 2:00am Eastern time
Facility: Tor2
Affected Device(s): dist04.tor2, core04.tor2
Affected Customers: Customers who are directly connected to the affected devices above.
Service Impacting: Yes - Customers physically connected to the affected devices above will see services impacted during this maintenance window. This maintenance activity is not expected to significantly impact customers whose equipment is not directly connected.

Details: Before the maintenance work commences, we will be steering all non-local traffic away from the devices in question, by gracefully shutting down BGP sessions, and adjusting metrics to push bits over other links during the maintenance period, to ensure impact is as limited as possible. We do not expect customers in the Tor1 facility to see any discernible impact from this maintenance period.

We will be replacing the memory in both supervisor cards (active/standby) in the core04.tor2 core router. During this operation, the router's OS will also be reloaded/rebooted in order to enable the Cisco NSF (Non-Stop Forwarding), and NSR (Non-Stop Routing) feature sets, which should allow the router to fail-over between its active/standby supervisors without dropping BGP sessions/links/etc.

On dist04.tor2 we will be performing the same procedure, but removing the additional memory from the platform (as the distribution layer has very little prospect of actually using the additional memory, which has proven to be more of a liability than benefit). Aside from that, the procedure will be the same as the core04.tor2 router (and we will likely to choose to work on the dist04.tor2 distribution router first, given it's position in the network/workflow).

After the configuration changes have been applied to both devices, we intend to simulate the active supervisor card failing in each device to ensure the configuration applied is suitable, and understand the specific impact of having the chassis failing from active to standby supervisor.

While the maintenance window in question is approximately 3 hours long to allow staff time to stage/prep/test/etc., we estimate the actual service impact will be somewhere between 10 and 20 minutes total per device over the course of 1-3 short interruptions.