autonat: add metrics #2086
autonat: add metrics #2086
Conversation
|
@marten-seemann I've added. |
There was a problem hiding this comment.
ipfs/kubo#9647 was merged, so I could run this on my Kubo node. I had to rebase this branch on top of the current go-libp2p master first though.
This looks mostly good to me. I'm wondering if we should export the timestamp of the next probe (as calculated by scheduleProbe as well), might be nice to display that in Grafana as well.
| "refId": "A" | ||
| } | ||
| ], | ||
| "title": "Reachability status confidence", |
There was a problem hiding this comment.
Can you set the min and max values in the dashboard settings? Then this will be correctly scaled.
| "text": "Public" | ||
| }, | ||
| "2": { | ||
| "color": "red", |
There was a problem hiding this comment.
I'd prefer to use a different color here. Red could be interpreted as "something is going wrong", which is not necessarily the case if you're actually behind a NAT. Maybe orange or violet would be better?
There was a problem hiding this comment.
that's fair. violet seems good to me.
| "refId": "A" | ||
| } | ||
| ], | ||
| "title": "Server: Dial requests rate limited", |
There was a problem hiding this comment.
This doesn't seem to work?
There was a problem hiding this comment.
I think that's because no request is being rate limited. On my kubo node I was receiving very few requests. But I'll test this again.
There was a problem hiding this comment.
Just checked my dashboard, and now it's showing one event (after a couple of hours), so it seems to work.
Why is this not part of the dashboard to the right though?
There was a problem hiding this comment.
The dashboard on the right has segregation by Message_ResponseStatus. So this point would be counted under dial refused. It does seem fine to separate rate limit and dial refused and display on the same dashboard. Should I do that?
There was a problem hiding this comment.
Added other Refusal reasons too.
| "refId": "A" | ||
| } | ||
| ], | ||
| "title": "server: DialResponse By Type", |
There was a problem hiding this comment.
I have to admit, I'm always confused who's the client and server when thinking about AutoNAT. Maybe "Outgoing Dials" would be harder to mix up?
There was a problem hiding this comment.
outgoing does seem better. will fix.
sukunrt
force-pushed
the
autonat-metrics
branch
from
27e14ff
to
fe66815
Compare
fixes: #2017