multipleof4/ntfy
1
0
Fork 0
mirror of https://github.com/multipleof4/ntfy.git synced 2026-10-08 21:05:21 +00:00
Code Issues Packages Projects Releases Wiki Activity

WIP: Clsuter support (cross node delivery, leader election)

Browse Source
This commit is contained in:
binwiederhier
2026-08-01 16:10:31 +02:00
parent dc11655153
commit 5a6c4277ad
35 changed files with 3232 additions and 63 deletions
Download Patch File Download Diff File Expand all files Collapse all files
server/server.yml
+28
View File
@@ -61,6 +61,34 @@
#
# database-url: <connection-string>
# If "cluster-listen" is set, clustering is implicitly enabled: this node registers itself in
# the PostgreSQL node registry and fans published messages out to the other cluster nodes over
# HTTP, so subscribers connected to any node receive messages published to any other node.
# Requires "database-url" and "cluster-secret".
#
# - cluster-listen is the ip:port of the dedicated fan-out listener that peer nodes talk to.
# Bind it to a private network interface (e.g. "10.0.0.5:2587"); the public listeners never
# serve the fan-out endpoint.
# - cluster-node-id is a stable per-node identifier (e.g. the hostname); required.
# - cluster-advertise-url is the base URL peer nodes use to reach this node's fan-out listener;
# it defaults to "http://<cluster-listen>". It must be set explicitly if cluster-listen binds
# a wildcard address (e.g. ":2587").
# - cluster-secret authenticates node-to-node fan-out requests; it must be identical on all
# nodes.
# - cluster-batch-linger is how long fan-out messages may wait to form a batch per peer node,
# trading up to that much cross-node delivery latency for a bounded request rate between
# nodes. Set to 0 to send each message immediately.
#
# SECURITY: The fan-out endpoint injects messages into arbitrary topics. The shared secret
# protects it, and it is only served on the dedicated cluster listener -- but you should still
# make sure that listener is reachable only from the private network (firewall/VPC rules).
#
# cluster-listen: <ip:port>
# cluster-node-id: <hostname>
# cluster-advertise-url: "http://<cluster-listen>"
# cluster-secret: <secret>
# cluster-batch-linger: 500ms
# If "cache-file" is set, messages are cached in a local SQLite database instead of only in-memory.
# This allows for service restarts without losing messages in support of the since= parameter.
# Not required if "database-url" is set (messages are stored in PostgreSQL instead).