v0.3.0

[andydunstall]

v0.3.0 contains a rewrite of the agent to server protocol. Instead of forwarding individual HTTP requests and responses, the protocol transports raw TCP bytes multiplexed over a single outbound connection from the upstream to the server.

This has a few major benefits:

• It supports non-HTTP protocols in the future (particularly TCP)
• It supports WebSockets
• It supports a Go SDK in the future where clients can listen for connections from Piko like any other TCP listener (such as net.Listener in Go)
• It makes the agent and server much simpler as they can treat the 'connection' to the upstream as any other TCP connection (even though the connection is actually multiplexed over the upstreams outbound connection to Piko)

This release also adds TLS support, both for the server and for the agent to configure custom root CAs to verify the servers TLS certificates. It also refactors the piko agent command line and configuration.

What's Changed

• docs: add logo by @andydunstall in docs: add logo #46
• client: add go sdk interface by @andydunstall in client: add go sdk interface #47
• agentv2: add agent cli and configuration by @andydunstall in agentv2: add agent cli and configuration #48
• agentv2: add endpoint interface by @andydunstall in agentv2: add endpoint interface #49
• agentv2: add endpoint reverse proxy by @andydunstall in agentv2: add endpoint reverse proxy #50
• conn: add websocket options by @andydunstall in conn: add websocket options #52
• tls: add tls support by @andydunstall in tls: add tls support #53
• agentv2: add heartbeats by @andydunstall in agentv2: add heartbeats #54
• docs: update logo by @andydunstall in docs: update logo #55
• client: register endpoint listener by @andydunstall in client: register endpoint listener #56
• serverv2: add reverse proxy to forward requests to upstream by @andydunstall in serverv2: add reverse proxy to forward requests to upstream #57
• agentv2: handle listener connections by @andydunstall in agentv2: handle listener connections #58
• add docker package by @andydunstall in add docker package #59
• tests: use httptest.Server by @andydunstall in tests: use httptest.Server #60
• serverv2: add reverse proxy by @andydunstall in serverv2: add reverse proxy #61
• agentv2: add pkg/mux by @andydunstall in agentv2: add pkg/mux #62
• agentv2: add connection per listener by @andydunstall in agentv2: add connection per listener #63
• agentv2: add config options by @andydunstall in agentv2: add config options #64
• serverv2: rewrite reverse proxy by @andydunstall in serverv2: rewrite reverse proxy #65
• serverv2: load balance among connected upstreams by @andydunstall in serverv2: load balance among connected upstreams #66
• piko: integrate agent/server v2 by @andydunstall in piko: integrate agent/server v2 #67
• admin: add request forwarding by @andydunstall in admin: add request forwarding #68
• status: add upstream endpoints status command by @andydunstall in status: add upstream endpoints status command #69
• server: add usage reporting by @andydunstall in server: add usage reporting #70

Full Changelog: v0.2.0...v0.3.0

---

This discussion was created from the release v0.3.0.

[randalee]

I upgraded the Helm-based deployment on EKS from version 0.2 to 0.3. I simply changed the provided image tag, but the pod failed to start.

{"level":"info","ts":"2024-06-12T01:29:39.709Z","subsystem":"main","msg":"starting piko server","node-id":"piko-0-lrwuftl","version":"v0.3.0"}
{"level":"info","ts":"2024-06-12T01:29:39.709Z","subsystem":"gossip","msg":"starting gossip","node-id":"piko-0-lrwuftl","bind-addr":":8003","advertise-addr":"172.20.6.189:8003"}
{"level":"warn","ts":"2024-06-12T01:29:39.716Z","subsystem":"main","msg":"failed to join cluster","error":"resolve: piko:8003: lookup host: piko: lookup piko on 10.100.0.10:53: no such host"}
{"level":"info","ts":"2024-06-12T01:29:39.717Z","subsystem":"proxy","msg":"starting proxy server","addr":"[::]:8000"}
{"level":"info","ts":"2024-06-12T01:29:39.717Z","subsystem":"admin","msg":"starting admin server","addr":"[::]:8002"}
{"level":"info","ts":"2024-06-12T01:29:39.717Z","subsystem":"admin","msg":"starting upstream server","addr":"[::]:8001"}
{"level":"warn","ts":"2024-06-12T01:29:40.761Z","subsystem":"gossip","msg":"failed to join cluster","error":"resolve: piko:8003: lookup host: piko: lookup piko on 10.100.0.10:53: no such host"}
{"level":"warn","ts":"2024-06-12T01:29:43.031Z","subsystem":"gossip","msg":"failed to join cluster","error":"resolve: piko:8003: lookup host: piko: lookup piko on 10.100.0.10:53: no such host"}

In the case of version 0.2, there are no issues on EKS, and for version 0.3, there are no problems when running it locally with Docker.

 docker run ghcr.io/andydunstall/piko:v0.3.0 server
{"level":"info","ts":"2024-06-12T01:50:33.555Z","subsystem":"main","msg":"starting piko server","node-id":"pbkwevb","version":"v0.3.0"}
{"level":"info","ts":"2024-06-12T01:50:33.555Z","subsystem":"gossip","msg":"starting gossip","node-id":"pbkwevb","bind-addr":":8003","advertise-addr":"10.10.0.2:8003"}
{"level":"info","ts":"2024-06-12T01:50:33.556Z","subsystem":"admin","msg":"starting upstream server","addr":"[::]:8001"}
{"level":"info","ts":"2024-06-12T01:50:33.556Z","subsystem":"admin","msg":"starting admin server","addr":"[::]:8002"}
{"level":"info","ts":"2024-06-12T01:50:33.556Z","subsystem":"proxy","msg":"starting proxy server","addr":"[::]:8000"}

[andydunstall]

Hi @randalee, sorry I removed the /ready route by mistake, so if the pod never becomes ready it won't be added to the service so DNS never resolves

I've merged a short term fix, and will have a look at improving how joining a cluster works in K8S next

[andydunstall]

I've released v0.4.0 with the fix so helm should now work, you'll still see failed to join cluster logs in the first pod as it takes a few seconds for the DNS record to be created by K8S, but it should eventually succeed

[randalee]

I've released v0.4.0 with the fix so helm should now work, you'll still see failed to join cluster logs in the first pod as it takes a few seconds for the DNS record to be created by K8S, but it should eventually succeed

Thank you for your prompt feedback. Version 0.4 works well on EKS! 💯