May is conference season. Kicking off the season, I gave a short presentation last week at a STAC research event in London. As you can imagine, it is no surprise that AI has arrived in the finance industry. An angle I was unaware of before and only realized at the conference is that what we usually call AI today can also be seen as a method to approximate arbitrary functions (in a mathematical sense). So, when you can do inference with a known latency, you train your model on an arbitrary trading algorithm and use the (hardware-accelerated) inference for your speed-trading production decisions. Buy, sell, or hold. Neat.
The other conference that LINBIT is attending this month is the Red Hat Summit in Denver. While I was unable this year, I have been to several summits in years past. However, my US colleagues were there to experience the excitement of the show. Unsurprisingly, a glance at the program reveals that virtually everybody has connected their topic to AI in one way or another.
By the way, Linus Torvalds, as we know him, is not overly impressed by AI. Expressing skepticism, he made headlines in 2015 when he stated that people who believe in AI leading a singularity event must be on drugs. His recent comments on the topic indicate that AI improves Linux’s Nvidia driver situation.
Over at LINBIT, a week after Fedora 40, Ubuntu 24.04 (“Noble”), and Proxmox adopted the Linux-6.8.0 kernel, I was relieved that we had the DRBD release ready in time. It only brought fixes, and no one reported a regression in the last week.
Also worth noting is that before the drbd-9.2.9 release, we had to keep some Kubernetes customers on drbd-9.1.x because they triggered a hard-to-reproduce bug. We fixed the bug, which the customers have confirmed. With the release of drbd-9.2.9 we recommend every DRBD user to leave drbd-9.1.x or older behind.
Over to content, ‘Monitoring & Performing Actions on DRBD Resources in Real-Time Using DRBDmon’ is a blog post that outlines how DRBDmon brings DRBD resource monitoring and maintenance convenience to high-availability storage clusters. The blog post links to a DRBDmon walk-through video presented by the utility’s lead developer, Robert Altnoeder, which I will link here for your convenience.
We also published ‘Using the Prometheus Operator for Monitoring LINBIT SDS in Kubernetes.’ Within the post, we demonstrate how LINBIT SDS makes it simple to get started monitoring and alerting significant LINSTOR and DRBD events.
Last but certainly not least, ‘Open Source VMware Alternatives That Can Integrate With LINBIT Storage Solutions’ is our latest response to the ongoing VMware/Broadcom story, which I have covered in previous editions of this newsletter.
Over to recent software releases, of which there have been quite a few.
linstor-server 1.27.1 introduces support for java-21, which is now default shipped with ubuntu-noble. There are also some fixes for zfs 2.2 and user set drbd verify-algorithm.
LINSTOR Operator v2.5.1 contains updated LINSTOR, CSI driver, and DRBD versions, as well as default priority class settings for all components.
We also published drbd-utils v9.28.0. We did not find any bugs since the last RC, so the release shares the text from the announcement mail of drbd-utils v9.28.0-rc.1.
linstor-proxmox v8.0.2 is a small bugfix release for the version 8 series of the plugin.
drbd-reactor v1.4.1 includes various smaller fixes. The main reason for the release is that reactor benefits from the changes that occurred in drbd-utils (out of order netlink message handling). This release allows us to bump the dependency on drbd-utils to the latest version.