KinesisRecord make_kinesis_record(int min_serialized_size = 0)
{
var kr = new KinesisRecord();
int num_user_records = 1 + (new Random().Next() % 100);
Func <UserRecord> make_ur = () =>
{
return(TestUtil.make_user_record(
new Random().Next().ToString(),
new Random().Next().ToString(),
new Random().Next().ToString(),
10000,
"myStream"));
};
for (int i = 0; i < num_user_records; i++)
{
kr.add(make_ur());
}
if (min_serialized_size > 0)
{
while ((int)kr.accurate_size() < min_serialized_size)
{
kr.add(make_ur());
}
}
return(kr);
}
public static void verify(List <UserRecord> original, KinesisRecord kr)
{
AggregatedRecord ar = null;
verify_format(original, kr, ref ar);
verify_content(original, ar);
}
public void KinesisRecordUnitTest_Clearing()
{
int N = 10;
List <UserRecord> user_records = new List <UserRecord>();
KinesisRecord r = new KinesisRecord();
for (int i = 0; i < N; i++)
{
var ur = TestUtil.make_user_record();
user_records.Add(ur);
r.add(ur);
}
TestUtil.verify(user_records, r);
for (int i = 0; i < 5; i++)
{
r.clear();
user_records.Clear();
for (int j = 0; j < N; j++)
{
var ur = TestUtil.make_user_record();
user_records.Add(ur);
r.add(ur);
}
TestUtil.verify(user_records, r);
}
}
public void KinesisRecordUnitTest_AccurateSize()
{
for (int i = 0; i < 100; i++)
{
KinesisRecord r = new KinesisRecord();
long num_records = new Random().Next(1, 512);
for (long j = 0; j < num_records; j++)
{
int key_size = new Random().Next(1, 256);
int data_size = new Random().Next(1, 64 * 1024);
var ur = TestUtil.make_user_record(new string('a', key_size),
new string('a', data_size),
new Random().Next() % 2 == 0 ? "" : "123");
r.add(ur);
}
int predicted = (int)r.accurate_size();
string serialized = r.serialize();
if (r.Items().Count > 1)
{
Assert.AreEqual(predicted, serialized.Length - 16);
}
else
{
Assert.AreEqual(predicted, serialized.Length);
}
}
}
public void ReducerUnitTest_CountLimit()
{
int limit = 100;
var reducer = make_reducer(int.MaxValue, limit);
List <UserRecord> v = new List <UserRecord>();
for (int j = 0; j < 3; j++)
{
v.Clear();
for (int i = 0; i < limit; i++)
{
var ur = TestUtil.make_user_record();
v.Add(ur);
KinesisRecord result = reducer.add(ur) as KinesisRecord;
if (i < limit - 1)
{
Assert.IsNull(result);
}
else
{
Assert.IsNotNull(result);
TestUtil.verify(v, result);
Assert.AreEqual((int)reducer.size(), 0);
}
}
}
}
KinesisRecord make_kinesis_record(DateTime deadline, DateTime expiration)
{
var ur = TestUtil.make_user_record();
ur.set_deadline(deadline);
ur.set_expiration(expiration);
ur.Predicted_shard(0);
var kr = new KinesisRecord();
kr.add(ur);
return(kr);
}
public void KinesisRecordUnitTest_EstimatedSize()
{
for (int i = 0; i < 50; i++)
{
KinesisRecord r = new KinesisRecord();
int num_records = new Random().Next(2, 512);
// non repeated partition keys
for (int j = 0; j < num_records; j++)
{
int key_size = new Random().Next(1, 256);
int data_size = new Random().Next(1, 64 * 1024);
var ur = TestUtil.make_user_record(new string('a', key_size),
new string('a', data_size),
new Random().Next() % 2 == 0 ? "" : "123");
r.add(ur);
}
// repeated partition keys
int key_size1 = new Random().Next(1, 256);
for (long j = 0; j < num_records; j++)
{
int data_size = new Random().Next(1, 64 * 1024);
var ur = TestUtil.make_user_record(new string('a', key_size1),
new string('a', data_size),
new Random().Next() % 2 == 0 ? "" : "123");
r.add(ur);
}
// small keys small data
for (long j = 0; j < num_records; j++)
{
var ur = TestUtil.make_user_record(new string('a', 2),
new string('a', 2),
new Random().Next() % 2 == 0 ? "" : "123");
r.add(ur);
}
int estimated = (int)r.Estimated_size();
string serialized = r.serialize();
double diff = (double)serialized.Length - estimated;
double percentage_diff = diff / serialized.Length * 100;
percentage_diff *= percentage_diff < 0 ? -1 : 1;
StringBuilder ss = new StringBuilder();
ss.Append("Estimated size should be within 1 percent or 32 bytes of actual ")
.Append("size, estimate was ").Append(estimated).Append(", actual size was ")
.Append(serialized.Length).Append(" (").Append(percentage_diff).Append("% difference)");
//BOOST_CHECK_MESSAGE(percentage_diff < 1 || diff < 32, ss.ToString());
}
}
public static void verify_unaggregated(UserRecord ur, KinesisRecord kr)
{
var serialized = kr.serialize();
Assert.AreEqual(ur.Data().ToString(Encoding.Default), serialized);
Assert.AreEqual(ur.Partition_key(), kr.partition_key());
if (ur.explicit_hash_key().ToString() != "-1")
{
Assert.AreEqual(ur.explicit_hash_key().ToString(), kr.explicit_hash_key());
}
else
{
Assert.AreEqual(KPLNETInterface.Utils.GetDecimalHashKey(ur.Partition_key()).ToString(), kr.explicit_hash_key());
}
}
public void ReducerUnitTest_NonEmpty()
{
KinesisRecord kr = null;
var reducer = make_reducer(
int.MaxValue,
int.MaxValue,
(result) =>
{
kr = result;
});
reducer.Flush();
Thread.Sleep(100);
Assert.IsNull(kr); //"Flush should not have produced a KinesisRecord because the Reducer is empty"
}
public void KinesisRecordUnitTest_Empty()
{
KinesisRecord r = new KinesisRecord();
Assert.AreEqual(0, (int)r.accurate_size());
Assert.AreEqual(0, (int)r.Estimated_size());
try
{
r.serialize();
Assert.Fail("Calling serialize on empty KinesisRecord should cause exception");
}
catch (Exception e)
{
// expected
}
}
public void KinesisRecordUnitTest_SingleRecordTestMethod()
{
// No explicit hash key (ehk)
{
KinesisRecord r = new KinesisRecord();
var ur = TestUtil.make_user_record();
r.add(ur);
TestUtil.verify_unaggregated(ur, r);
}
// With ehk
{
KinesisRecord r = new KinesisRecord();
var ur = TestUtil.make_user_record("a", "a", "123");
r.add(ur);
TestUtil.verify_unaggregated(ur, r);
}
}
public void ReducerUnitTest_Deadline()
{
KinesisRecord kr = null;
var reducer = make_reducer(
int.MaxValue,
int.MaxValue,
(result) => { kr = result; });
List <UserRecord> v = new List <UserRecord>();
for (int i = 0; i < 100; i++)
{
var ur = TestUtil.make_user_record(
"pk",
new Random().Next().ToString(),
"123",
5000 + i);
v.Add(ur);
reducer.add(ur);
}
// Should not flush after 1 second because we've set the deadlines to be 5
Thread.Sleep(1000);
Assert.AreEqual((int)reducer.size(), 100);
// Now we're going to put a record with a deadline that's now to trigger the
// flush.
var ur1 = TestUtil.make_user_record(
"pk",
new Random().Next().ToString(),
"123",
0);
reducer.add(ur1);
// This record should be moved to the front when flushed because of its
// deadline
v.Insert(0, ur1);
Thread.Sleep(300);
Assert.AreEqual((int)reducer.size(), 0);
Assert.IsNotNull(kr);
TestUtil.verify(v, kr);
}
AwsKinesisResult make_prr_ctx(int num_kr, int num_ur_per_kr, string error, Amazon.Kinesis.Model.PutRecordsResponse outcome, DateTime start, DateTime end)
{
List <KinesisRecord> krs = new List <KinesisRecord>();
for (int i = 0; i < num_kr; i++)
{
var kr = new KinesisRecord();
for (int j = 0; j < num_ur_per_kr; j++)
{
var ur = TestUtil.make_user_record();
ur.Predicted_shard(i);
kr.add(ur);
}
krs.Add(kr);
}
AwsKinesisResult result = new AwsKinesisResult(error, new AwsKinesisResponse(outcome), new PutRecordsRequest(), start, end);
result.context <PutRecordsRequest>().Items().AddRange(krs);
return(result);
}
public static void verify_format(List <UserRecord> original, KinesisRecord kr, ref AggregatedRecord container)
{
container = null;
byte[] serialized = kr.SerializedAggregatedRecord;
// verify magic number
byte[] expected_magic = null;
unchecked
{
expected_magic = new byte[] { (byte)-13, (byte)-119, (byte)-102, (byte)-62 };
}
int magic_len = expected_magic.Length;
byte[] magic = serialized.Take(magic_len).ToArray();
Assert.IsTrue(KPLNETInterface.Utils.AreArrayEqual(expected_magic, magic));
// verify protobuf payload
byte[] payload = serialized.Skip(expected_magic.Length).Take(serialized.Length - 16 - magic_len).ToArray();
container = AggregatedRecord.Parser.ParseFrom(Google.Protobuf.ByteString.CopyFrom(payload));
Assert.IsNotNull(container);
// verify md5 checksum
Assert.IsTrue(KPLNETInterface.Utils.AreArrayEqual(KPLNETInterface.Utils.GetMD5(payload), serialized.Skip(serialized.Length - 16).Take(16).ToArray()));
// verify the explicit hash key set on the Kinesis record
List <string> acceptable_hash_keys = new List <string>();
foreach (var ur in original)
{
if (ur.explicit_hash_key() > -1)
{
acceptable_hash_keys.Add(ur.explicit_hash_key().ToString());
}
else
{
acceptable_hash_keys.Add(KPLNETInterface.Utils.GetDecimalHashKey(ur.Partition_key()).ToString());
}
}
Assert.IsTrue(acceptable_hash_keys.Exists((i) => i == kr.explicit_hash_key()));
}
public void KinesisRecordUnitTest_SameEHK()
{
int N = 1000;
int M = 100;
List <UserRecord> user_records = new List <UserRecord>();
KinesisRecord r = new KinesisRecord();
for (int i = 0; i < N; i++)
{
var ur = TestUtil.make_user_record("pk", "data", "123");
user_records.Add(ur);
r.add(ur);
}
for (int i = 0; i < M; i++)
{
r.remove_last();
user_records.RemoveAt(user_records.Count - 1);
}
TestUtil.verify(user_records, r);
}
public void KinesisRecordUnitTest_Deadlines()
{
// The nearest deadline should always be kept
{
KinesisRecord r = new KinesisRecord();
var start = DateTime.Now;
for (int i = 0; i < 10; i++)
{
var ur = TestUtil.make_user_record();
ur.set_deadline(start.AddMilliseconds(i * 100));
ur.set_expiration(start.AddMilliseconds(i * 100));
r.add(ur);
}
Assert.IsTrue(r.Deadline() == start);
Assert.IsTrue(r.Expiration() == start);
}
// If a nearer deadline comes in, it should override the previous
{
KinesisRecord r = new KinesisRecord();
var earlier = DateTime.Now;
var later = earlier.AddMilliseconds(500);
{
var ur = TestUtil.make_user_record();
ur.set_deadline(later);
r.add(ur);
Assert.IsTrue(r.Deadline() == later);
}
{
var ur = TestUtil.make_user_record();
ur.set_deadline(earlier);
r.add(ur);
Assert.IsTrue(r.Deadline() == earlier);
}
// Removing the last added record should restore the previous deadline
r.remove_last();
Assert.IsTrue(r.Deadline() == later);
}
}
public void KinesisRecordUnitTest_DifferentParitionKey()
{
int N = 1000;
int M = 100;
List <UserRecord> user_records = new List <UserRecord>();
KinesisRecord r = new KinesisRecord();
for (int i = 0; i < N; i++)
{
var ur = TestUtil.make_user_record(i.ToString());
user_records.Add(ur);
r.add(ur);
}
for (int i = 0; i < M; i++)
{
r.remove_last();
user_records.RemoveAt(user_records.Count - 1);
}
TestUtil.verify(user_records, r);
}
public void AggregatorUnitTest_BasicTestMethod()
{
var aggregator = make_aggregator();
for (int shard_id = 1; shard_id <= 3; shard_id++)
{
List <UserRecord> v = new List <UserRecord>();
KinesisRecord kr = null;
for (int i = 0; i < kCountLimit; i++)
{
var ur = TestUtil.make_user_record("pk", TestUtil.random_string(new Random().Next(100)), TestUtil.get_hash_key(shard_id).ToString(), 10000 + i, "MyStream", 0);
kr = aggregator.put(ur);
v.Add(ur);
Assert.IsFalse(ur.Predicted_shard() == -1);
Assert.AreEqual(ur.Predicted_shard(), shard_id);
}
Assert.IsNotNull(kr);
TestUtil.verify(v, kr);
}
}
public void KinesisRecordUnitTest_MixedParitionKey()
{
List <string> keys = new List <string>()
{
"a", "b", "b", "a", "a", "c", "b", "a", "d", "e", "e", "d", "d", "d"
};
List <UserRecord> user_records = new List <UserRecord>();
KinesisRecord r = new KinesisRecord();
for (int i = 0; i < keys.Count; i++)
{
var ur = TestUtil.make_user_record(keys[i]);
user_records.Add(ur);
r.add(ur);
}
for (int i = 0; i < 3; i++)
{
r.remove_last();
user_records.RemoveAt(user_records.Count - 1);
}
TestUtil.verify(user_records, r);
}
public void KinesisRecordUnitTest_MixedEHK()
{
List <string> keys = new List <string>()
{
"1", "2", "2", "1", "1", "3", "2", "1", "4", "5", "5", "4", "4", "4"
};
List <UserRecord> user_records = new List <UserRecord>();
KinesisRecord r = new KinesisRecord();
for (int i = 0; i < keys.Count; i++)
{
var ur = TestUtil.make_user_record("pk", "data", keys[i]);
user_records.Add(ur);
r.add(ur);
}
for (int i = 0; i < 3; i++)
{
r.remove_last();
user_records.RemoveAt(user_records.Count - 1);
}
TestUtil.verify(user_records, r);
}
public void ReducerUnitTest_ResetTimeout()
{
KinesisRecord kr = null;
var reducer = make_reducer(
300,
int.MaxValue,
(result) =>
{
kr = result;
});
List <UserRecord> v = new List <UserRecord>();
int k = 0;
for (int i = 0; i < 10; i++)
{
var ur = TestUtil.make_user_record(
"pk",
new Random().Next().ToString(),
"123",
2000 + k++);
v.Add(ur);
reducer.add(ur);
}
// This record has a much closer deadline
{
var ur = TestUtil.make_user_record(
"pk",
new Random().Next().ToString(),
"123",
100);
v.Insert(0, ur);
reducer.add(ur);
}
// Put records until flush happens
KinesisRecord kr2;
{
do
{
// The other records have a further deadline.
var ur = TestUtil.make_user_record(
"pk",
new Random().Next().ToString(),
"123",
2000 + k++);
v.Add(ur);
kr2 = reducer.add(ur) as KinesisRecord;
} while (kr2 == null);
}
// Put a few more to make sure the buffer is not empty
for (int i = 0; i < 10; i++)
{
var ur = TestUtil.make_user_record(
"pk",
new Random().Next().ToString(),
"123",
2000 + k++);
v.Add(ur);
reducer.add(ur);
}
// No flush should happen in the next second even though we had a record with
// a 100ms deadline - that record should've gotten flushed by the size limit,
// and its deadline should no longer apply.
Thread.Sleep(1000);
Assert.IsNull(kr);
// The timer should now be set to the min deadline of the remaining records.
// It should go off after another second or so
Thread.Sleep(1500);
Assert.IsNotNull(kr);
// Check data integrity
Assert.AreEqual(kr.Size() + kr2.Size() + reducer.size(), (ulong)v.Count);
for (int i = v.Count - 1; i >= (int)kr.Size() + (int)kr2.Size(); i--)
{
v.RemoveAt(i);
}
List <UserRecord> v2 = new List <UserRecord>();
for (int i = (int)kr2.Size() - 1; i >= 0; i--)
{
v2.Insert(0, v[i]);
v.RemoveAt(i);
}
TestUtil.verify(v, kr);
TestUtil.verify(v2, kr2);
}
public void ReducerUnitTest_Concurrency()
{
int counter = 0;
//LOG(info) << "Starting concurrency test. If this doesn't finish in 30 " << "seconds or so it probably means there's a deadlock.";
ConcurrentBag <UserRecord> put = new ConcurrentBag <UserRecord>();
ConcurrentBag <KinesisRecord> results = new ConcurrentBag <KinesisRecord>();
var reducer = make_reducer(5000, int.MaxValue, (result) => { results.Add(result); });
List <Thread> threads = new List <Thread>();
for (int i = 0; i < 16; i++)
{
var t = new Thread(() =>
{
for (int j = 0; j < 32; j++)
{
KinesisRecord kr = null;
do
{
int count = Interlocked.Increment(ref counter);
var ur = TestUtil.make_user_record("pk", count.ToString(), "123", count);
put.Add(ur);
kr = reducer.add(ur) as KinesisRecord;
} while (kr == null);
results.Add(kr);
// Call flush sometimes too to mix things up further
if (i % 8 == 0 && i == j)
{
reducer.Flush();
}
}
});
t.Start();
threads.Add(t);
}
foreach (var t in threads)
{
t.Join();
}
while (reducer.size() > 0)
{
reducer.Flush();
Thread.Sleep(1000);
}
Assert.AreEqual((int)reducer.size(), 0);
//LOG(info) << "Finished putting data, " << counter << " records put. "
// << "Analyzing results...";
// Check that all records made it out. Order is no longer guaranteed with
// many workers, but we've put a monotonically increasing number in the
// record data, so that will allow us to sort and match records up.
List <UserRecord> put_v = new List <UserRecord>(put.ToArray());
List <UserRecord> result_v = new List <UserRecord>();
var resultsArray = results.ToArray();
for (int i = 0; i < results.Count; i++)
{
result_v.AddRange(resultsArray[i].Items());
}
Comparison <UserRecord> ckr = (a, b) => { return(string.Compare(a.Data().ToStringUtf8(), b.Data().ToStringUtf8())); };
put_v.Sort(ckr);
result_v.Sort(ckr);
Assert.AreEqual(put_v.Count, result_v.Count);
for (int i = 0; i < put_v.Count; i++)
{
Assert.AreEqual(put_v[i], result_v[i]);
}
}
