Graph with no explicit maxpoll (no pps either)

Hal Murray hmurray at
Sun Oct 30 03:13:44 UTC 2016

gem at said:
>> The server was alive and well during the whole time.  Most of the
>> time it was polling at 1024 seconds.
> Since the new plot joins point 1024 or less seconds apart something else is
> going on.  Clearly most of your data is spread out by more than 1024
> seconds. 

That was what I was trying to tell you a few days ago.

There is a filter between the raw packets and peerstats.  Only a few of the 
packets make it through.  loopstats is only updated when peerstats for 
sys_peer gets updated.

>> dig through the log files if it matters.)
> Can you just send them to me?

OK.  Sent off list.

> Well, it sure looks missing to me.  Those are pretty sparse dots.

The system was working as expected.

> You can
> adjust the 1024 to a value of your choosing.  Just look in the plot_slice()
> function.  Can you suggest a value that looks good to you, without being
> misleading? 

You can either make the 1024 in your test big enough to cover the normal 
case, or you can pick a number like an hour or two and live with the false 

The last column in loopstats is the (log 2 of the) polling interval.  You 
could multiply that by 8 on the fly.

These are my opinions.  I hate spam.

More information about the devel mailing list