[RPKI] Routinator 0.8.2-rc1 released

[Chris Caputo]

On Mon, 7 Dec 2020, Martin Hoffmann via RPKI wrote:
> Chris Caputo via RPKI wrote:
> > With 0.8.1:
> > 
> >    1618 afrinic
> >   62023 apnic
> >   19149 arin
> >   11915 lacnic
> >  110711 ripe
> > 
> > With 0.8.2-rc1:
> > 
> >    1618 afrinic
> >   62531 apnic
> >   19149 arin
> >   11917 lacnic
> >  110711 ripe
> > 
> > The lacnic change may just have been an update, while the apnic count 
> > change is dramatic.
> 
> I had a quick look at our own graphs. Currently they show
> 62,506 v. 62,589 VRPs for the APNIC region (These include duplicates;
> there seem to be about 80 duplicates). This difference has been 83 VRPs
> ever since I updated one server to the release candidate on Friday.
> 
> It seems correct, too: looking at the log, there are three CAs in the
> CNNIC repository that have ROAs rejected which would have been dropped
> fully under the 0.8.1 rules. These three CAs produce 83 VRPs in the
> 0.8.2 data set.
> 
> I’m not sure where your much lower number for 0.8.1 comes from. If you
> still have the log, perhaps you can have a look? Maybe some updates
> failed temporarily? 
> 
> > Unsure if related, but I noticed that first-run (empty ~/.rpki-cache)
> > took about 4 minutes with 0.8.1 but with 0.8.2-rc1 a first run took
> > ~12.5 minutes.  I emptied ~/.rpki-cache again and a second fresh
> > attempt took ~9.25 minutes.  (subsequent runs with a non-fresh
> > ~/.rpki-cache took anywhere from 67 to 72 seconds.)
>
> Update times seem to exhibit quite a bit of fluctuation, even with
> cache present. The graph shows one update that took 1200 seconds on one
> of our test machines.
>
> So I _think_ you just hit a somewhat bad time when updating? The code
> for fetching hasn’t changed at all.
>
> Can you perchance try again to confirm? If you are running Routinator
> in server mode, you can also get metrics for the update times for the
> individual repository servers to find the one the slows everything down.

I performed the following in order, with fresh compiles...

Fresh 0.8.1 .cargo and .rpki-cache directories.  Initial run:

  $ ~/.cargo/bin/routinator --version                                            
  Routinator 0.8.1                                                                         
  
  $ ~/.cargo/bin/routinator init --accept-arin-rpa

  $ time ~/.cargo/bin/routinator --verbose vrps --format csv --output /tmp/rpki-export-routinator.csv.tmp
  real    2m15.727s
  user    0m22.495s
  sys     0m6.748s

   1629 afrinic
  62485 apnic
  19238 arin
  11975 lacnic
 110744 ripe

Fresh 0.8.2-rc1 .cargo and .rpki-cache directories.  Initial run:

  $ ~/.cargo/bin/routinator --version                                                                                                                                    
  Routinator 0.8.2-rc1                                                                     

  $ ~/.cargo/bin/routinator init --accept-arin-rpa

  $ time ~/.cargo/bin/routinator --verbose vrps --format csv --output /tmp/rpki-export-routinator.csv.tmp
  real    3m27.682s
  user    0m23.926s
  sys     0m6.993s

   1629 afrinic
  62569 apnic
  19240 arin
  11976 lacnic
 110753 ripe

0.8.1 second run with saved 0.8.1 .rpki-cache dir:

  $ ~/.cargo/bin/routinator --version                                            
  Routinator 0.8.1                                                                         

  $ time ~/.cargo/bin/routinator --verbose vrps --format csv --output /tmp/rpki-export-routinator.csv.tmp
  real    1m14.051s
  user    0m24.398s
  sys     0m7.803s

   1629 afrinic
  62486 apnic
  19240 arin
  11976 lacnic
 110753 ripe

0.8.2-rc1 second run with saved 0.8.2-rc1 .rpki-cache dir:

  $ ~/.cargo/bin/routinator --version                                                                                                                                    
  Routinator 0.8.2-rc1                                                                     

  $ time ~/.cargo/bin/routinator --verbose vrps --format csv --output /tmp/rpki-export-routinator.csv.tmp
  real    0m57.267s
  user    0m17.879s
  sys     0m4.999s

   1629 afrinic
  62569 apnic
  19240 arin
  11976 lacnic
 110753 ripe

The 83 or 84 difference in APNIC numbers for 0.8.1 vs 0.8.2-rc1, appears 
to match what you have also observed.  I am not sure about that larger 
difference I previously observed.  I assume it was a unique issue at the 
time.

Thanks,
Chris