24th, Christmas: BIND slaves decide to suddenly stop accepting zone transfers from the master. Half a day of raging and I still couldn't figure out why. dig axfr works fine, but the slaves refuse a zone update according to tcpdump logs.

25th, 2nd day: A server decides to go down and take half my network with it. Turns out that a Python script managed to crash the goddamn kernel.

Thank you very much technology for making the Christmas days just a little bit better ❤️

At least I didn't have anything to do during either days, because of the COVID-19 pandemic. And to be fair, I did manage to make a Telegram bot with fancy webhooks and whatnot in 5MB of memory and 18MB of storage. Maybe I should just write the whole thing and make another sacred temple where shitty code gets beaten the fuck out of the system. Terry must've been onto something...

  • 5
    Sounds like a fun holiday!

    Also, check your dimms :) when userspace crashes kernelspace [esp in innocent ways], and especially when random magic starts happening all over the place - it's about time to consider dimm replacement...
  • 3
    All hail TempleOS.
  • 4
    It's your fault for using the oppressive term "slaves". If you had called them "workers" or "agents" they would have had better self esteem and accepted zone transfers.

    This is entirely your fault tbh.

  • 1
    @junon Hah! NGL after looking at my BIND configs for both for a while, I legit thought that a few times. In the DNS they renamed those to primary/secondary now. RFC8499 is the one to look at for this. ISC in its quest to be as RFC compliant as possible, immediately implemented it... Also ISC is more or less the biggest IETF member in the DNS scene. Paul Vixie - founder of ISC - also wanted to call it initiator/responder.. confused yet? There was a whole naming contest when I brought it up on the DNSOP mailing list, and as you could probably imagine, I was immediately labeled a racist for wanting to get back to master/slave :')

    Oh well.. Good thing that I did all 3 DNS servers (1 hidden master, 2 public slaves) complete with zones and configs in git, with a branch for each. So I was able to just check out the slaves' repos to that of the master and restart the service, as if nothing happened. Still gotta figure out why though... That master really shouldn't be as useless as it is now.
  • 1
    @Condor > configs in git

    Hopefully they're on the main branch, not the master branch. That's the next place I'd check.

Add Comment