Changing leader after initialization? IE, what if the leader disconnects? about rqlite HOT 5 CLOSED

When you say "change the leader" you mean the cluster elects a new leader, right? I just want to be sure.

If I understand this issue, it's basically the same as wanting redirects, right? If we had that, this wouldn't be an issue, correct?

from rqlite.

The cluster now has basic leader redirection, though I have not performed extensive testing. I just checked that curl -L .... is redirected to the leader node. Let me know how it goes.

from rqlite.

When you say "change the leader" you mean the cluster elects a new leader, right? I just want to be sure.

Yep that's correct.

I just did some testing and it looks like things are okay! I set up the 3 node example, took down node.1(the first leader). And the outputs look good.

Node 3 was arbitarily chosen as the new leader when I did this:

curl localhost:4002/raft?pretty
{
    "leader": "1d4e168",
    "name": "7d2502e",
    "peers": {
        "1b2bcd5": {
            "name": "1b2bcd5",
            "connectionString": "http://localhost:4001"
        },
        "1d4e168": {
            "name": "1d4e168",
            "connectionString": "http://localhost:4003"
        }
    },
    "state": "follower"
}
 curl localhost:4003/raft?pretty
{
    "leader": "1d4e168",
    "name": "1d4e168",
    "peers": {
        "1b2bcd5": {
            "name": "1b2bcd5",
            "connectionString": "http://localhost:4001"
        },
        "7d2502e": {
            "name": "7d2502e",
            "connectionString": "http://localhost:4002"
        }
    },
    "state": "leader"

One minor issue though, it doesn't look like the -L parameter works with the raft endpoint. Node 3 is the leader in my case, but the raft endpoint for node 2 still shows node 2's data.

curl -L localhost:4002/raft?pretty
{
    "leader": "1d4e168",
    "name": "7d2502e",
    "peers": {
        "1b2bcd5": {
            "name": "1b2bcd5",
            "connectionString": "http://localhost:4001"
        },
        "1d4e168": {
            "name": "1d4e168",
            "connectionString": "http://localhost:4003"
        }
    },
    "state": "follower"

IMO, this doesn't really matter because the same node's raft endpoint tells you enough data anyway.

Looking back over my first post in this thread, I'm seeing that I made an error and that rqlite is actually doing the leader reselection perfectly.

You can close this out if you like. Remember to add the -L page to your Readme. Thanks a ton man, this stuff is pretty exciting for me.

from rqlite.

Let me look into the issue you mentioned above, but that is a good idea about adding -L to the README.

from rqlite.

Ok, yeah, I only put leader redirection in place for the write endpoint. If you want stats, diags, etc, you need to go out and hit each node individually. Let me know if this is an issue.

from rqlite.