Reported Outages

Resolved: Work is complete and KVM@TACC is available.

On Wednesday, March 27 2024 KVM@TACC will be unavailable between 8:00am CDT and 4:00pm CDT to perform necessary system maintenance. Access to the KVM@TACC web interface and API will be unavaiable, and network access to instances will be interrupted. Instances will continue to run, even though they will not be reachable.

5:00 PM We have a workaround in place that should restore connectivity to all floating IPs, however we are still investigating the root cause. There may be additonal minor interruptions while we troubleshoot, but currently all floating IPs tested respond to ping and ssh.

1:30 PM: The maintenance is done, but we are receiving reports of failure to SSH to public IPs from for some users and some instances, and are investigating. Access to the API and dashboard have been fully restored.

Update 5:45 PM Jan 31: Jupyterhub and CHI@Edge are back online.

This morning the VM cluster that hosts Portal, CHI@Edge, and JupyterHub went down. This affected the ability to create new leases at any Chameleon site, and access to the help desk.

Update: 3pm Feb 8: Outage is resolved, all skylake and rtx_6000 nodes now have 2 10G network interfaces (up from one each)

Update: 7PM Feb 6th: The site is online, and all P3 nodes (everything except compute_skylake and gpu_rtx_6000) are accessible.

On Feb 6th, 2024, we'll be taking the CHI@UC site down for maintenance, in order to replace some failing network hardware.

CHI@NCAR will be down for maintenance from January 26th to January 29th inclusive, in order to facilitate work on the datacenter racks currently hosting it.

Update: The upgrade is complete, and Jupyter service should be as normal

---

We will be upgrading JupyterHub versions, which will take down JupyterHub and running JupyterLab instances for a short duration.

2:22 PM: Issues affecting Jupyterhub are resolved.

The root cause was a network interruption affecting availability of the storage services which Juptyerhub's infra depends on. After the network interruption was resolved, dependent services, including Jupyterhub, returned to a healthy state.

We are still investigating other issues affecting KVM@TACC, unrelated to this outage. If your work in Jupyter depends on KVM, you may be able to continue your work by using a different Chameleon site.

Dec 13: The outage is now resolved.

You should observe that I/O performance on ephemeral disks is back to normal, as they have been migrated to a different storage backend.
Note: If you attach an additonal cinder volume of type "ceph-hdd", that volume may experience unexpectedly delayed I/O as the system state settles, but this will not impact the majority of instances.

We're observing intermittent issues accessing Chameleon's Jupyterhub. There was a failure in one of the VMs underlying it, and in combination, some requests are still being routed to the failed instance in error.

In some cases, users may be able to work around this error by refreshing their browser.

Site staff are investigating to revive both the VM, and to correct the request routing.

The outage is now resolved.

Due to a combination of factors, a misconfigured node was able to cause a DoS of control plane services. The relevant traffic has been blocked, and the site is back online.