We have finished updating all storage servers with new Linux kernels and Ceph updates as planned. We have seen intermittent reduced performance as expected.
The cluster is currently finishing up some additional replication tasks that aren't affecting performance.
Posted 9 days ago. Oct 15, 2017 - 13:32 CEST
Scheduled maintenance is currently in progress. We will provide updates as necessary.
Posted 9 days ago. Oct 15, 2017 - 08:00 CEST
We are going to upgrade our Ceph cluster to the newest stable version with our release branch (Hammer) and will also upgrade to the latest kernel release in our long term release series (4.9). This fixes a few known issues in our underlying filesystem (XFS) and prepares our cluster for a larger Ceph version update in Q4 (Jewel).
This requires a reboot of all storage servers, which will be applied in a slow and staggered fashion to reduce potential impact. After our previous improvements, including our network optimizations, you might not see any impact of the recoveries. Nevertheless, as we have not yet seen a fully silent maintenance in production we're announcing this maintenance with the expectation of up to 1 minute of reduced storage performance for each of our 10 servers.