mtbc | Unplanned remote server maintenance

Prompted by the need to patch for Meltdown, my KVM-based virtual hosting provider effected some kind of Linux update since which OpenBSD guest instances are behaving intolerably. There is probably some issue with the paravirtualized run-time clock implementation: an instance that I rely on now suffers dreadful clock drift and exhibits strangeness like,

$ time sleep 1
    0m40.50s real     0m00.00s user     0m00.01s system

I could wait for the problem to be fixed but my hosting provider cannot easily apply workaround command-line options to particular nodes. OpenBSD folks believe that the cause is KVM's emulation of local APIC in repeated mode, expressing the reasonable attitude that, Working around bugs in someone else's software certainly isn't enough motivation for me to implement it. My obvious choices are:

Stick with my hosting provider. They are generally good and this problem is not their fault so better the devil I know. A switch to NetBSD would take some work but may be sufficiently palatable even if it moves me away from LibreSSL. If the switch goes easily, in the longer term I could migrate my other machines and enjoy a happier life through no longer reading the harsh incivility from OpenBSD's founder to others.

Online (and who else?) offer trivially cheap Dedibox BareMetal Servers that may allow me to run OpenBSD while avoiding hypervisor bugs and Meltdown-like vulnerabilities to others' instances. I could directly port much of my existing setup to the new system and may eventually save a little money overall.