Ticket #1392: test_get_latencies_upgraded01.darcs.2.patch

Tue Apr 26 14:59:58 MDT 2011  wilcoxjg@gmail.com
  * test_storage.py:  test_latencies now expects None in output categories that contain too few samples for the associated percentile to be unambiguously reported.

Tue Apr 26 15:16:41 MDT 2011  wilcoxjg@gmail.com
  * server.py:  get_latencies now reports percentiles _only_ if there are sufficient observations for the interpretation of the percentile to be unambiguous.

New patches:

[test_storage.py:  test_latencies now expects None in output categories that contain too few samples for the associated percentile to be unambiguously reported.
wilcoxjg@gmail.com**20110426205958
 Ignore-this: 2cf1920eb878f97394940584c470f43a
] {
hunk ./src/allmydata/test/test_storage.py 1314
             ss.add_latency("allocate", 1.0 * i)
         for i in range(1000):
             ss.add_latency("renew", 1.0 * i)
+        for i in range(20):
+            ss.add_latency("write", 1.0 * i)
         for i in range(10):
             ss.add_latency("cancel", 2.0 * i)
         ss.add_latency("get", 5.0)
hunk ./src/allmydata/test/test_storage.py 1323
         output = ss.get_latencies()
 
         self.failUnlessEqual(sorted(output.keys()),
-                             sorted(["allocate", "renew", "cancel", "get"]))
+                             sorted(["allocate", "renew", "cancel", "write", "get"]))
         self.failUnlessEqual(len(ss.latencies["allocate"]), 1000)
         self.failUnless(abs(output["allocate"]["mean"] - 9500) < 1, output)
         self.failUnless(abs(output["allocate"]["01_0_percentile"] - 9010) < 1, output)
hunk ./src/allmydata/test/test_storage.py 1344
         self.failUnless(abs(output["renew"]["99_0_percentile"] - 990) < 1, output)
         self.failUnless(abs(output["renew"]["99_9_percentile"] - 999) < 1, output)
 
+        self.failUnlessEqual(len(ss.latencies["write"]), 20)
+        self.failUnless(abs(output["write"]["mean"] - 9) < 1, output)
+        self.failUnless(output["write"]["01_0_percentile"] == None, output)
+        self.failUnless(abs(output["write"]["10_0_percentile"] -  2) < 1, output)
+        self.failUnless(abs(output["write"]["50_0_percentile"] - 10) < 1, output)
+        self.failUnless(abs(output["write"]["90_0_percentile"] - 18) < 1, output)
+        self.failUnless(abs(output["write"]["95_0_percentile"] - 19) < 1, output)
+        self.failUnless(output["write"]["99_0_percentile"] == None, output)
+        self.failUnless(output["write"]["99_9_percentile"] == None, output)
+
         self.failUnlessEqual(len(ss.latencies["cancel"]), 10)
         self.failUnless(abs(output["cancel"]["mean"] - 9) < 1, output)
hunk ./src/allmydata/test/test_storage.py 1356
-        self.failUnless(abs(output["cancel"]["01_0_percentile"] -  0) < 1, output)
+        self.failUnless(output["cancel"]["01_0_percentile"] == None, output)
         self.failUnless(abs(output["cancel"]["10_0_percentile"] -  2) < 1, output)
         self.failUnless(abs(output["cancel"]["50_0_percentile"] - 10) < 1, output)
         self.failUnless(abs(output["cancel"]["90_0_percentile"] - 18) < 1, output)
hunk ./src/allmydata/test/test_storage.py 1360
-        self.failUnless(abs(output["cancel"]["95_0_percentile"] - 18) < 1, output)
-        self.failUnless(abs(output["cancel"]["99_0_percentile"] - 18) < 1, output)
-        self.failUnless(abs(output["cancel"]["99_9_percentile"] - 18) < 1, output)
+        self.failUnless(output["cancel"]["95_0_percentile"] == None, output)
+        self.failUnless(output["cancel"]["99_0_percentile"] == None, output)
+        self.failUnless(output["cancel"]["99_9_percentile"] == None, output)
 
         self.failUnlessEqual(len(ss.latencies["get"]), 1)
hunk ./src/allmydata/test/test_storage.py 1365
-        self.failUnless(abs(output["get"]["mean"] - 5) < 1, output)
-        self.failUnless(abs(output["get"]["01_0_percentile"] - 5) < 1, output)
-        self.failUnless(abs(output["get"]["10_0_percentile"] - 5) < 1, output)
-        self.failUnless(abs(output["get"]["50_0_percentile"] - 5) < 1, output)
-        self.failUnless(abs(output["get"]["90_0_percentile"] - 5) < 1, output)
-        self.failUnless(abs(output["get"]["95_0_percentile"] - 5) < 1, output)
-        self.failUnless(abs(output["get"]["99_0_percentile"] - 5) < 1, output)
-        self.failUnless(abs(output["get"]["99_9_percentile"] - 5) < 1, output)
+        self.failUnless(output["get"]["mean"] == None, output)
+        self.failUnless(output["get"]["01_0_percentile"] == None, output)
+        self.failUnless(output["get"]["10_0_percentile"] == None, output)
+        self.failUnless(output["get"]["50_0_percentile"] == None, output)
+        self.failUnless(output["get"]["90_0_percentile"] == None, output)
+        self.failUnless(output["get"]["95_0_percentile"] == None, output)
+        self.failUnless(output["get"]["99_0_percentile"] == None, output)
+        self.failUnless(output["get"]["99_9_percentile"] == None, output)
 
 def remove_tags(s):
     s = re.sub(r'<[^>]*>', ' ', s)
}
[server.py:  get_latencies now reports percentiles _only_ if there are sufficient observations for the interpretation of the percentile to be unambiguous.
wilcoxjg@gmail.com**20110426211641
 Ignore-this: 546001f34d53e35ce2025b05b4ea66b6
] {
hunk ./src/allmydata/storage/server.py 119
 
     def get_latencies(self):
         """Return a dict, indexed by category, that contains a dict of
-        latency numbers for each category. Each dict will contain the
+        latency numbers for each category. If there are sufficient samples 
+        for unambiguous interpretation, each dict will contain the
         following keys: mean, 01_0_percentile, 10_0_percentile,
         50_0_percentile (median), 90_0_percentile, 95_0_percentile,
hunk ./src/allmydata/storage/server.py 123
-        99_0_percentile, 99_9_percentile. If no samples have been collected
-        for the given category, then that category name will not be present
-        in the return value."""
+        99_0_percentile, 99_9_percentile.  If there are insufficient 
+        samples for a given percentile to be interpreted unambiguously
+        that percentile will be reported as None. If no samples have been 
+        collected for the given category, then that category name will 
+        not be present in the return value. """
         # note that Amazon's Dynamo paper says they use 99.9% percentile.
         output = {}
         for category in self.latencies:
hunk ./src/allmydata/storage/server.py 135
                 continue
             stats = {}
             samples = self.latencies[category][:]
-            samples.sort()
             count = len(samples)
hunk ./src/allmydata/storage/server.py 136
-            stats["mean"] = sum(samples) / count
-            stats["01_0_percentile"] = samples[int(0.01 * count)]
-            stats["10_0_percentile"] = samples[int(0.1 * count)]
-            stats["50_0_percentile"] = samples[int(0.5 * count)]
-            stats["90_0_percentile"] = samples[int(0.9 * count)]
-            stats["95_0_percentile"] = samples[int(0.95 * count)]
-            stats["99_0_percentile"] = samples[int(0.99 * count)]
-            stats["99_9_percentile"] = samples[int(0.999 * count)]
+            stats["samplesize"] = count
+            samples.sort()
+            if count > 1:
+                stats["mean"] = sum(samples) / count
+            else:
+                stats["mean"] = None
+
+            orderstatlist = [(0.01, "01_0_percentile", 100), (0.1, "10_0_percentile", 10),\
+                             (0.50, "50_0_percentile", 10), (0.90, "90_0_percentile", 10),\
+                             (0.95, "95_0_percentile", 20), (0.99, "99_0_percentile", 100),\
+                             (0.999, "99_9_percentile", 1000)]
+
+            for percentile, percentilestring, minnumtoobserve in orderstatlist: 
+                if count >= minnumtoobserve:
+                    stats[percentilestring] = samples[int(percentile*count)]
+                else:
+                    stats[percentilestring] = None
+
             output[category] = stats
         return output
 
}

Context:

[corrected "k must never be smaller than N" to "k must never be greater than N"
secorp@allmydata.org**20110425010308
 Ignore-this: 233129505d6c70860087f22541805eac
] 
[Fix a test failure in test_package_initialization on Python 2.4.x due to exceptions being stringified differently than in later versions of Python. refs #1389
david-sarah@jacaranda.org**20110411190738
 Ignore-this: 7847d26bc117c328c679f08a7baee519
] 
[tests: add test for including the ImportError message and traceback entry in the summary of errors from importing dependencies. refs #1389
david-sarah@jacaranda.org**20110410155844
 Ignore-this: fbecdbeb0d06a0f875fe8d4030aabafa
] 
[allmydata/__init__.py: preserve the message and last traceback entry (file, line number, function, and source line) of ImportErrors in the package versions string. fixes #1389
david-sarah@jacaranda.org**20110410155705
 Ignore-this: 2f87b8b327906cf8bfca9440a0904900
] 
[remove unused variable detected by pyflakes
zooko@zooko.com**20110407172231
 Ignore-this: 7344652d5e0720af822070d91f03daf9
] 
[allmydata/__init__.py: Nicer reporting of unparseable version numbers in dependencies. fixes #1388
david-sarah@jacaranda.org**20110401202750
 Ignore-this: 9c6bd599259d2405e1caadbb3e0d8c7f
] 
[update FTP-and-SFTP.rst: the necessary patch is included in Twisted-10.1
Brian Warner <warner@lothar.com>**20110325232511
 Ignore-this: d5307faa6900f143193bfbe14e0f01a
] 
[control.py: remove all uses of s.get_serverid()
warner@lothar.com**20110227011203
 Ignore-this: f80a787953bd7fa3d40e828bde00e855
] 
[web: remove some uses of s.get_serverid(), not all
warner@lothar.com**20110227011159
 Ignore-this: a9347d9cf6436537a47edc6efde9f8be
] 
[immutable/downloader/fetcher.py: remove all get_serverid() calls
warner@lothar.com**20110227011156
 Ignore-this: fb5ef018ade1749348b546ec24f7f09a
] 
[immutable/downloader/fetcher.py: fix diversity bug in server-response handling
warner@lothar.com**20110227011153
 Ignore-this: bcd62232c9159371ae8a16ff63d22c1b
 
 When blocks terminate (either COMPLETE or CORRUPT/DEAD/BADSEGNUM), the
 _shares_from_server dict was being popped incorrectly (using shnum as the
 index instead of serverid). I'm still thinking through the consequences of
 this bug. It was probably benign and really hard to detect. I think it would
 cause us to incorrectly believe that we're pulling too many shares from a
 server, and thus prefer a different server rather than asking for a second
 share from the first server. The diversity code is intended to spread out the
 number of shares simultaneously being requested from each server, but with
 this bug, it might be spreading out the total number of shares requested at
 all, not just simultaneously. (note that SegmentFetcher is scoped to a single
 segment, so the effect doesn't last very long).
] 
[immutable/downloader/share.py: reduce get_serverid(), one left, update ext deps
warner@lothar.com**20110227011150
 Ignore-this: d8d56dd8e7b280792b40105e13664554
 
 test_download.py: create+check MyShare instances better, make sure they share
 Server objects, now that finder.py cares
] 
[immutable/downloader/finder.py: reduce use of get_serverid(), one left
warner@lothar.com**20110227011146
 Ignore-this: 5785be173b491ae8a78faf5142892020
] 
[immutable/offloaded.py: reduce use of get_serverid() a bit more
warner@lothar.com**20110227011142
 Ignore-this: b48acc1b2ae1b311da7f3ba4ffba38f
] 
[immutable/upload.py: reduce use of get_serverid()
warner@lothar.com**20110227011138
 Ignore-this: ffdd7ff32bca890782119a6e9f1495f6
] 
[immutable/checker.py: remove some uses of s.get_serverid(), not all
warner@lothar.com**20110227011134
 Ignore-this: e480a37efa9e94e8016d826c492f626e
] 
[add remaining get_* methods to storage_client.Server, NoNetworkServer, and
warner@lothar.com**20110227011132
 Ignore-this: 6078279ddf42b179996a4b53bee8c421
 MockIServer stubs
] 
[upload.py: rearrange _make_trackers a bit, no behavior changes
warner@lothar.com**20110227011128
 Ignore-this: 296d4819e2af452b107177aef6ebb40f
] 
[happinessutil.py: finally rename merge_peers to merge_servers
warner@lothar.com**20110227011124
 Ignore-this: c8cd381fea1dd888899cb71e4f86de6e
] 
[test_upload.py: factor out FakeServerTracker
warner@lothar.com**20110227011120
 Ignore-this: 6c182cba90e908221099472cc159325b
] 
[test_upload.py: server-vs-tracker cleanup
warner@lothar.com**20110227011115
 Ignore-this: 2915133be1a3ba456e8603885437e03
] 
[happinessutil.py: server-vs-tracker cleanup
warner@lothar.com**20110227011111
 Ignore-this: b856c84033562d7d718cae7cb01085a9
] 
[upload.py: more tracker-vs-server cleanup
warner@lothar.com**20110227011107
 Ignore-this: bb75ed2afef55e47c085b35def2de315
] 
[upload.py: fix var names to avoid confusion between 'trackers' and 'servers'
warner@lothar.com**20110227011103
 Ignore-this: 5d5e3415b7d2732d92f42413c25d205d
] 
[refactor: s/peer/server/ in immutable/upload, happinessutil.py, test_upload
warner@lothar.com**20110227011100
 Ignore-this: 7ea858755cbe5896ac212a925840fe68
 
 No behavioral changes, just updating variable/method names and log messages.
 The effects outside these three files should be minimal: some exception
 messages changed (to say "server" instead of "peer"), and some internal class
 names were changed. A few things still use "peer" to minimize external
 changes, like UploadResults.timings["peer_selection"] and
 happinessutil.merge_peers, which can be changed later.
] 
[storage_client.py: clean up test_add_server/test_add_descriptor, remove .test_servers
warner@lothar.com**20110227011056
 Ignore-this: efad933e78179d3d5fdcd6d1ef2b19cc
] 
[test_client.py, upload.py:: remove KiB/MiB/etc constants, and other dead code
warner@lothar.com**20110227011051
 Ignore-this: dc83c5794c2afc4f81e592f689c0dc2d
] 
[test: increase timeout on a network test because Francois's ARM machine hit that timeout
zooko@zooko.com**20110317165909
 Ignore-this: 380c345cdcbd196268ca5b65664ac85b
 I'm skeptical that the test was proceeding correctly but ran out of time. It seems more likely that it had gotten hung. But if we raise the timeout to an even more extravagant number then we can be even more certain that the test was never going to finish.
] 
[docs/configuration.rst: add a "Frontend Configuration" section
Brian Warner <warner@lothar.com>**20110222014323
 Ignore-this: 657018aa501fe4f0efef9851628444ca
 
 this points to docs/frontends/*.rst, which were previously underlinked
] 
[web/filenode.py: avoid calling req.finish() on closed HTTP connections. Closes #1366
"Brian Warner <warner@lothar.com>"**20110221061544
 Ignore-this: 799d4de19933f2309b3c0c19a63bb888
] 
[Add unit tests for cross_check_pkg_resources_versus_import, and a regression test for ref #1355. This requires a little refactoring to make it testable.
david-sarah@jacaranda.org**20110221015817
 Ignore-this: 51d181698f8c20d3aca58b057e9c475a
] 
[allmydata/__init__.py: .name was used in place of the correct .__name__ when printing an exception. Also, robustify string formatting by using %r instead of %s in some places. fixes #1355.
david-sarah@jacaranda.org**20110221020125
 Ignore-this: b0744ed58f161bf188e037bad077fc48
] 
[Refactor StorageFarmBroker handling of servers
Brian Warner <warner@lothar.com>**20110221015804
 Ignore-this: 842144ed92f5717699b8f580eab32a51
 
 Pass around IServer instance instead of (peerid, rref) tuple. Replace
 "descriptor" with "server". Other replacements:
 
  get_all_servers -> get_connected_servers/get_known_servers
  get_servers_for_index -> get_servers_for_psi (now returns IServers)
 
 This change still needs to be pushed further down: lots of code is now
 getting the IServer and then distributing (peerid, rref) internally.
 Instead, it ought to distribute the IServer internally and delay
 extracting a serverid or rref until the last moment.
 
 no_network.py was updated to retain parallelism.
] 
[TAG allmydata-tahoe-1.8.2
warner@lothar.com**20110131020101] 
Patch bundle hash:
b0cd7b8e35f8f9ccb234a36671df7ce51cb1c8e5