Https redirect to http protocol

[weenxin]

I am using service discovery deployed in localhost.

you can think there is an dns routine:

node0 : rqlite-quickstart-0.rqlite-quickstart-gvr.demo
node1 : rqlite-quickstart-1.rqlite-quickstart-gvr.demo
node2 : rqlite-quickstart-2.rqlite-quickstart-gvr.demo

and I deploy node0 using this command line

rqlited -http-addr 0.0.0.0:4001 -raft-addr 0.0.0.0:4002 \ 
-http-adv-addr rqlite-quickstart-0.rqlite-quickstart-gvr.demo:4001 \
-raft-adv-addr rqlite-quickstart-0.rqlite-quickstart-gvr.demo:4002 \
-disco-id 0a863868-38d6-46bb-b8b1-ff8a0c1189f0 \
-disco-url http://rqlite-service-discovery.rqlite:8090 \
-http-key /data/configdb/rqlite.pem \
-http-cert /data/configdb/client.pem \
-http-no-verify 
-node-encrypt 
-node-key /data/configdb/rqlite.pem \
-node-cert /data/configdb/client.pem \
-node-no-verify \
/data/db

node0 work well.

and then I start my second and third node using

rqlited -http-addr 0.0.0.0:4001 -raft-addr 0.0.0.0:4002 \ 
-http-adv-addr rqlite-quickstart-1.rqlite-quickstart-gvr.demo:4001 \
-raft-adv-addr rqlite-quickstart-1.rqlite-quickstart-gvr.demo:4002 \
-disco-id 0a863868-38d6-46bb-b8b1-ff8a0c1189f0 \
-disco-url http://rqlite-service-discovery.rqlite:8090 \
-http-key /data/configdb/rqlite.pem \
-http-cert /data/configdb/client.pem \
-http-no-verify 
-node-encrypt 
-node-key /data/configdb/rqlite.pem \
-node-cert /data/configdb/client.pem \
-node-no-verify \
/data/db

rqlited -http-addr 0.0.0.0:4001 -raft-addr 0.0.0.0:4002 \ 
-http-adv-addr rqlite-quickstart-2.rqlite-quickstart-gvr.demo:4001 \
-raft-adv-addr rqlite-quickstart-2.rqlite-quickstart-gvr.demo:4002 \
-disco-id 0a863868-38d6-46bb-b8b1-ff8a0c1189f0 \
-disco-url http://rqlite-service-discovery.rqlite:8090 \
-http-key /data/configdb/rqlite.pem \
-http-cert /data/configdb/client.pem \
-http-no-verify 
-node-encrypt 
-node-key /data/configdb/rqlite.pem \
-node-cert /data/configdb/client.pem \
-node-no-verify \
/data/db

But, Node1 , Node2 log is like

cluster-join] 2019/11/26 06:38:52 failed to join cluster at [rqlite-quickstart-0.rqlite-quickstart-gvr.demo:4001], sleeping 5s before retry
2019/11/26 06:38:54 [WARN] raft: EnableSingleNode disabled, and no known peers. Aborting election.
[cluster-join] 2019/11/26 06:38:57 failed to join cluster at [rqlite-quickstart-0.rqlite-quickstart-gvr.demo:4001], sleeping 5s before retry
[cluster-join] 2019/11/26 06:39:02 failed to join cluster at [rqlite-quickstart-0.rqlite-quickstart-gvr.demo:4001], sleeping 5s before retry
[cluster-join] 2019/11/26 06:39:07 failed to join cluster at [rqlite-quickstart-0.rqlite-quickstart-gvr.demo:4001], after 3 attempts
[rqlited] 2019/11/26 06:39:07 failed to join cluster at [rqlite-quickstart-0.rqlite-quickstart-gvr.demo:4001]: \
Post http://rqlite-quickstart-0.rqlite-quickstart-gvr.demo:4001/join: \
net/http: HTTP/1.x transport connection broken: malformed HTTP response "\x15\x03\x01\x00\x02\x02\x16"

Seems like rqlited using http schema rather than https to join this cluster.

[otoolep]

I can repro this issue with the in-cloud disco service. The issue is that the nodes are not storing enough information in the disco service entry -- they should also store the protocol the other nodes are using (HTTP or HTTPS).

This should be easy enough to fix, but I also forgot to version the data written by nodes, to the disco service. Changing the format of the entry up there would mean older versions couldn't use the disco service to join to clusters running older nodes. Perhaps it's not a big deal -- AWS monitoring shows that the disco service is not heavily used. People could just upgrade their clusters I guess.

[otoolep]

I am proposing this patch instead: #579

[otoolep]

This issue should be fixed on master now. Build master (you may need to reclone your repo since I force updated master) and try it out.

[otoolep]

Can see how v4.6.0 (just released) works for you?

[otoolep]

I believe this is effectively fixed by v4.6.0. Please re-open if not.