[Update – 00:31] The maintenance is complete for tonight. Core networking upgrades were completed, however individual top-of-rack switches were not moved to new hardware and will require future, planned maintenance. Please contact SDSC Operations with any questions or concerns.
[Update – 23:59] Project storage and Universal Scale Storage systems are online. Affected XSEDE systems are also online.
[Update – 23:14] There are remaining problems with some project storage node networking, as a result some NFS mounts may be unavailable. In addition, routing to some XSEDE resources may be affected at this time.
[Update – 21:14] Issues with the H23 switchgroup and core routing paths are being investigated. As a result, some systems may appear offline longer than originally expected.
[Update – 20:07] Core router hardware has been upgraded. New distribution switches are in place and downstream connected equipment is now being migrated over.
[Update – 19:04] Engineers have begun the maintenance.
—
We have rescheduled this maintenance for the evening of September 12th, beginning at 1900. We have also extended the window an extra hour to ensure there is time to complete all migrations. The scope and impacts remain the same.
SDSC engineers will perform a major upgrade to our data center aggregation-layer switches. We will be replacing our old Arista chassis switches with new Juniper hardware to accommodate demand for higher bandwidth connection in the research community. During the above stated window we will be migrating all connectivity coming into the data center core to this new hardware. This will cause multiple, brief network connectivity issues for all users of the SDSC address space, as well as the UCSD colo address space. Customer with direct external connectivity (via CENIC, etc.) will see no issues.
For context and planning, there will be one interruption shortly after 1900 when the uplinks to our router core are moved, and then over the course of the rest of the maintenance window customers will see another brief interruption as their top-of-rack switch, or direct connections are migrated to the new hardware.