parabo
hello
RN
Oye como va
parabo
RN
parabo
what is up?
dr|z3d
parabo: new I2P+ is up on postman.
dr|z3d
>>> I2P+ update now up on postman, please seed if you can, torrent is here: tracker2.postman.i2p/index.php?action=Download&id=62612 <<<
dr|z3d
>>> Details page: tracker2.postman.i2p/index.php?view=TorrentDetail&id=62612 <<<
parabo
hi
dr|z3d
what up p
not_bob
Too much :(
not_bob
But, that nasty i2pd bug has been fixed.
dr|z3d
you're alive!
not_bob
I am, and have been.
not_bob
You don't read my blog?
dr|z3d
not religiously. from time to time.
not_bob
Yeah, too much shit as usual.
not_bob
But, here's the good news. I can do a scan of all active eepsites in under 10 min!
dr|z3d
so term99 was telling ,e
not_bob
I think I could cut that in half too.
dr|z3d
*me
dr|z3d
except you have a problem, houston.
not_bob
Do tell.
not_bob
Am I not detecting new sites?
dr|z3d
you already know. load avg of 80...
not_bob
Huh?
not_bob
80?
dr|z3d
maybe I was misinformed.
not_bob
I'm not sure what you mean.
not_bob
I'm running under load avg of 16.
not_bob
Which is under 100%
dr|z3d
on how many cores?
not_bob
16
dr|z3d
ok
dr|z3d
that's fine, then.
not_bob
During a fast scan it may hit 14.
dr|z3d
well, in theory :)
not_bob
Yeah, it depends on how busy each core is, yeah.
not_bob
So, about that...
not_bob
I was having an issue with load spikes.
not_bob
But, I found a soltion to that.
dr|z3d
let's have a look at what your graphs are sating.
dr|z3d
*saying
dr|z3d
fugly as.
not_bob
Now I have a -d [delay] option to delay each ping's start by 1-15 seconds.
not_bob
So, they don't start all at the exact same time.
not_bob
Hmm?
dr|z3d
you see, there's the problem.. if you load the cores too heavily, then the ping results will be tainted.
not_bob
Yeah, I understand that.
dr|z3d
and if your graph for _your_ site is reporting a 6s+ delay, then you have a problem.
not_bob
It's nice to be able to do it *really fast*. But, I am aware of the other things.
not_bob
Oh, that's just magic.
not_bob
I'm sure it's never actually 6 seconds.
not_bob
I take the real ping time * 300 if it's my site.
dr|z3d
well, sure, the algo messes with the reported times on your site, but still. it shouldn't ever get there.
not_bob
It doens't.
not_bob
If you see 5 seconds for ping to my site, then it's 1/300th of that.
not_bob
I just turned it down to * 200
not_bob
We will see if that makes the graph look better for me.
dr|z3d
yeah, it's a bad way to calculate your own pings.
not_bob
I agree.
not_bob
I could show the real number, but then it would just be a flat line on that scale.
dr|z3d
better just to add a second to the reported results. you'll get a much easier to interpret graph that way.
dr|z3d
then any obvious spikes from the median will mean something.
dr|z3d
also useless.
dr|z3d
results + 1 second gives you something meaningful to look at.
not_bob
It would still be a flat line.
not_bob
A ping to my own site is measured in ms
dr|z3d
doesn't matter if it's flat, it's visible and any MAJOR deviation is instantly understood.
not_bob
Yeah, it would have to be pretty major.
not_bob
The last ping to my site was 18ms
not_bob
The previous one 12ms
dr|z3d
the other less instant fix is to put local pings on a different scale.
not_bob
That's a fiar point there.
not_bob
But, there is only one local ping, ever.
not_bob
So, I have been treating it as an edge case.
not_bob
And mostly ignoring it so long as it looks pretty :)
not_bob
Looking at my yearly graph, yeah, I see where it started to rise up.
not_bob
But, on the flip side. It shows that I've had pretty much 100% uptime.
not_bob
Did you see my end of year post on zzz about jumps? I appear to have handled about 1/3 of the i2p jumps in the last year.
dr|z3d
it's probably a good time to start thinking about a different default graph.
not_bob
Oh?
not_bob
Suggestions?
dr|z3d
yes, I saw the post.
not_bob
I could just have it pick a random site each time...
dr|z3d
nah, that's also not that interesting as a default. I'm thinking more a new graph type.
not_bob
Oh
dr|z3d
a graph or graphs that can fit in that space with stats, both relating to the sites listed and your site.
dr|z3d
pie graphs or wotnot.
not_bob
Ahh
not_bob
Now I understand.
not_bob
Yeah.
dr|z3d
successful/total jumps, visitors, whatever.
not_bob
Number of ties, percent stable and so on?
not_bob
Ahh!
not_bob
Yes, true.
not_bob
And, I do log all that data.
not_bob
Well, sorta. Visitors are hard to log. Unique visitors.
not_bob
Due to the way i2p works.
dr|z3d
and, sure, site-specific stuff, too, if there's space.
not_bob
Yeah
not_bob
Ok, I see what you mean.
not_bob
I'll see what I can do on that.
not_bob
Sketch a few things up, then try them.
not_bob
Though...
not_bob
I'm multi-homed.
dr|z3d
sure, it's a fun side-project.
not_bob
So, jumps and whatnot will never be right.
dr|z3d
maybe they will, when term99's finished with his fork :)
not_bob
If the main host creates the nifty graph, it would only unclude data from itself, not the daughter nodes.
not_bob
:)
not_bob
nodes/routers
dr|z3d
this is where a FLAT FILE db comes into play.
not_bob
zzz says I can have up to about 19 homes.
dr|z3d
something that can be shared between ping nodes.
not_bob
Yes
not_bob
He and I spoke about that.
not_bob
I've done quite a bit of thinking about it, but no time coding a solution.
not_bob
But, any solution I write would work with any kind of backend.
dr|z3d
um, no, you can't have 19 mult-homed routers.
not_bob
I can't?
not_bob
Was it 9?
dr|z3d
not unless you're running them all on 0 hop.
not_bob
How many can I have then?
not_bob
I read a post zzz wrote about it a while back.
dr|z3d
4 or 5 I believe is the limit.
not_bob
Oh
not_bob
Well, ok.
not_bob
I guess that's that then.
not_bob
In my testing I tend to get a new home every 10 min or so.
dr|z3d
if there's a case to be made for increasing the limit, talk to zzz in -dev at some point.
not_bob
Nah, I just want my jump service to be up 100% of the time.
dr|z3d
it's all about how many leasesets a floofill will accept for a single dest in a given timeframe (5m?)
not_bob
Ahh
not_bob
I was under the impression that the most recent leaseset won.
dr|z3d
when published to the same ff, sure.
not_bob
ok
dr|z3d
but there's this thing called a throttle.
not_bob
Ahh...
not_bob
Yeah, I don't want to hit that.
dr|z3d
keep publishing the same dest to the same ff, after a while it'll ignore you.
not_bob
Yeah
not_bob
I'm not to that point yet.
dr|z3d
more routers sharing the same dest, more chance of throttle.
not_bob
Good to know that 4 or 5 is the limit.
dr|z3d
anyways, with a flat file db you don't have to worry so much about uptime of individual ping nodes.
dr|z3d
you aggregate the data and discard the crap.
not_bob
remove any zeros, yeah.
not_bob
Saves disk space.
not_bob
And processing time.
not_bob
I could totally use an sql backend.
dr|z3d
it gives you the data in a way that makes it a lot more accessible for processing.
not_bob
Yep, I know sql
not_bob
And, I assume you are refering to a flat file version of that.
dr|z3d
note the FLAT FILE in the proposition. not a fully-fledged db backend.
not_bob
Ahh
not_bob
So, not like sqlite or the like.
dr|z3d
that second link is probably the most useful.
not_bob
Reading it now.
dr|z3d
it's not that far removed from what you're doing now, except instead of 1 file per feature per host, you put everything (related) in a single file.
not_bob
Yeah, I see that.
not_bob
And that was a path I thought about going down when I wrote it.
not_bob
But, went this way due to not having to worry about file locking and whatnot.
not_bob
But, yeah.
not_bob
Also, the way I do it means possibly less io/cpu to read data from just one site.
not_bob
But,yeah. I see the benifits too of going with a single flat file.
dr|z3d
flat file means you've got something easy to share around.
dr|z3d
something that can be pushed over rss or whatever.
not_bob
This is also true.
not_bob
Unrelated, do you have any idea what kinds of trouble little kids get into?
not_bob
Everything goes in the mouth.
not_bob
Alas, I need to wander off again.
not_bob
I'll be back eventually, don't worry.
dr|z3d
*** chuckles ***
dr|z3d
I2P+ I2PSnark standalone now up, please seed if you're able: tracker2.postman.i2p/index.php?action=Download&id=62638
parabo
hi