jschoubben
1f5b70a995
novox/hq ADR 0038. A module cannot choose a port: it is written once and assigned anywhere, so any number it picks is a guess about a machine it has never seen. A database module met the mesh's own store on 5432 and was told, by a container runtime three layers down, that the port was already allocated. The number used to appear three times in every module — the rule set, what a consumer is told, and what the runtime publishes — agreeing only because one person wrote all three. Now it appears once, in `listens`, and the other two are derived: the container publishes `20000:5432`, the consumer is told 20000, and the rule set opens 20000. An assignment is made once and kept, as a credential is. A port that moved on every declaration would restart both ends each time and hand a consumer a number that was true when it was read. Ports the protocol fixes — mail on 25, submission on 587, DNS on 53 — say so, and are then claims: one holder per machine, and the second is refused by name at assignment. That is the mechanism the mesh already has for what is singular on a machine, pointed at ports. A mapping written the long way is left exactly as it is. Some things must be pinned by hand, and quietly overruling somebody who wrote both halves would be worse than not offering the short form. Still open, and known: the substrate is not a module, so the mesh has never heard of its own store and cannot yet assign around it. That is what 028 will still be about after this.
51 lines
3.4 KiB
JSON
51 lines
3.4 KiB
JSON
{
|
|
"module": "dnsmasq",
|
|
"version": "1",
|
|
"requires": [
|
|
"resolver-data"
|
|
],
|
|
"provides": [
|
|
"wildcard-resolution"
|
|
],
|
|
"claims": [
|
|
{
|
|
"name": "the-dns-port",
|
|
"scope": "node"
|
|
}
|
|
],
|
|
"listens": [
|
|
{
|
|
"port": 53,
|
|
"protocol": "udp",
|
|
"from": "mesh",
|
|
"why": "names under every machine in this mesh, for this machine and what it runs",
|
|
"fixed": true
|
|
}
|
|
],
|
|
"resources": [
|
|
{
|
|
"id": "package",
|
|
"type": "package",
|
|
"package": "dnsmasq"
|
|
},
|
|
{
|
|
"id": "config",
|
|
"type": "file",
|
|
"path": "/etc/dnsmasq.conf",
|
|
"mode": "0644",
|
|
"content": "# Managed by the mesh. dnsmasq's own defaults are replaced whole rather than\n# patched, because this module owns the file and a patch would leave whatever\n# was there before to be discovered later.\n\n# What the mesh computed: one wildcard per machine, its name and everything\n# under it. Rewritten whenever a machine joins or leaves, which is why the\n# service below reflects it.\nconf-file=/etc/mesh-resolver/nodes.conf\n\n# Where it answers. Both are names the mesh chose, so this file needs to know\n# nothing about this particular machine:\n#\n# mesh0 the private network, so anything on it \u2014 including a container\n# on this machine \u2014 can ask.\n# 127.0.0.55 this machine's own use, for whatever points resolution at the\n# mesh. Not .53 or .54: systemd-resolved holds BOTH \u2014 .53 is its\n# stub and .54 its proxy stub \u2014 which this module asserted was\n# free until a machine said otherwise.\n#\n# Listening on a loopback address makes dnsmasq take the rest of\n# loopback with it, 127.0.0.1 included. That is why this module\n# claims `the-dns-port`: it takes the machine's DNS port, and\n# saying it takes only one address would be the same kind of\n# comfortable claim that .54 was free.\n#\n# .55 is a convention and not a reservation. If a future systemd\n# takes it, this line changes and nothing else does, which is the\n# reason it is written once here rather than in each module that\n# points at it.\n#\n# bind-dynamic rather than bind-interfaces: mesh0 does not exist until the\n# machine is on the private network, and binding an interface that is not there\n# yet fails to start rather than waiting for it.\nbind-dynamic\ninterface=mesh0\nlisten-address=127.0.0.55\n\n# **It forwards nothing, and must not read resolv.conf to find out where to.**\n# Whatever points this machine at the mesh writes its own address into\n# resolv.conf \u2014 so a resolver that read it for upstreams would find itself,\n# and every query it could not answer locally would loop until its receive\n# queue filled. That is not theoretical: it filled with 15KB of queries and\n# every lookup on the machine hung.\n#\n# It needs no upstream because it is never asked for anything else: the\n# asking module routes only the mesh's suffix here and leaves the rest\n# wherever the machine already sent it.\nno-resolv\ndomain-needed\nbogus-priv\n"
|
|
},
|
|
{
|
|
"id": "service",
|
|
"type": "service",
|
|
"unit": "dnsmasq.service",
|
|
"state": "running",
|
|
"boot": "enabled",
|
|
"restart-on": [
|
|
"config",
|
|
"mesh-resolver.nodes"
|
|
]
|
|
}
|
|
]
|
|
}
|