[tor-bugs] #18910 [Metrics/CollecTor]: distributing descriptors accross CollecTor instances

[Tor Bug Tracker & Wiki]

#18910: distributing descriptors accross CollecTor instances
-------------------------------+---------------------------------
 Reporter:  iwakeh             |          Owner:  iwakeh
     Type:  enhancement        |         Status:  needs_review
 Priority:  High               |      Milestone:  CollecTor 1.1.0
Component:  Metrics/CollecTor  |        Version:
 Severity:  Normal             |     Resolution:
 Keywords:  ctip               |  Actual Points:
Parent ID:                     |         Points:
 Reviewer:                     |        Sponsor:
-------------------------------+---------------------------------

Comment (by karsten):

 New bug, though I'm not certain yet where to fix it: the
 `DescriptorIndexCollector` instance of one module deletes the synchronized
 descriptors stored by the instance of another module.  Look at the
 following (new) code:

 {{{
     DescriptorCollector dc =
         DescriptorSourceFactory.createDescriptorCollector();
     dc.collectDescriptors("https://collector.torproject.org",
         new String[] { "/recent/torperf/" }, 0L, new File("sync"), true);
     dc.collectDescriptors("https://collector.torproject.org",
         new String[] { "/recent/exit-lists/" }, 0L, new File("sync"),
 true);
 }}}

 What would you expect that code to do?  Populate two local directories
 `sync/recent/torperf/` and `sync/recent/exit-lists/`?  Not quite, the last
 line clears all descriptors in `sync/recent/torperf/`.  Uhmmmmm.  And I'm
 not even sure what `DescriptorCollectorImpl` would do, possibly the same
 thing.

 We can either work around this in CollecTor by using separate
 subdirectories for the modules, like
 `sync/Relay/collector.torproject.org/recent/relay-descriptors/...`, or we
 can decide this is a (usability) bug in metrics-lib, put out a 1.5.1, and
 switch.

--
Ticket URL: <https://troodi.torproject.org/projects/tor/ticket/18910#comment:95>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online