**Yuki the Maven** @yuki_the_maven · Oct 22, 2017, 01:30

**Yuki the Maven** @yuki_the_maven · Oct 22, 2017, 01:30

Yuki the Maven @yuki_the_maven

> looks at pcap of http interactions
> response content encoding is gzipped
"/cry how do I see contents now"
> discover that wireshark automatically degzips
"omg great this made my day!"
> read the contents
> it says "No cluster leader"
ないねぇぇ
(pic: net-juu no susume)

**Yuki the Maven** @yuki_the_maven · Oct 22, 2017, 02:06

**Yuki the Maven** @yuki_the_maven · Oct 22, 2017, 02:06

great, so it's basically using the wrong interface to bind

**Yuki the Maven** @yuki_the_maven · 2017-10-22T03:17:06Z

Yuki the Maven @yuki_the_maven

so binding/advertising on 0.0.0.0 picks the first non-private interface, that for me is the wrong one, but fixing conf and restarting without nuking <data-dir> picks the wrong up again #nomad #cluster

I spent hours on something this stupid but this happens so often I'm not even pissed. I'll just go to bed and cry on my dakimakura. g'nite fediverse
(pic: net-juu no susume)

Oct 22, 2017, 03:17 · Web · 0 · 0

**Yuki the Maven** @yuki_the_maven · Oct 23, 2017, 02:37

**Yuki the Maven** @yuki_the_maven · Oct 23, 2017, 02:37

cluster ready. systemd unit fixed so it enables properly. `nomad plan` the task:
> Task Group: "patch-and-reboot" (3 create)
was expecting 1 as per update stanza… but ofc not because a one-time short-lived command is *not* an update. there is no max_parallel equivalent for create. total failure. all for nothing. and I could have noticed right away by paying more attention at the docs.

not. even. mad.
(pic: fate/apocrypha)