forked from thepaul/cassandra-dtest
-
Notifications
You must be signed in to change notification settings - Fork 0
/
incremental_repair_test.py
262 lines (200 loc) · 8.27 KB
/
incremental_repair_test.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
import os
import time
from re import findall
from unittest import skip
from cassandra import ConsistencyLevel
from nose.plugins.attrib import attr
from assertions import assert_almost_equal, assert_one
from ccmlib.node import Node
from dtest import Tester, debug
from tools import insert_c1c2, since
@since('2.1')
class TestIncRepair(Tester):
def __init__(self, *args, **kwargs):
kwargs['cluster_options'] = {'start_rpc': 'true'}
# Ignore these log patterns:
self.ignore_log_patterns = [
r'Can\'t send migration request: node.*is down',
]
Tester.__init__(self, *args, **kwargs)
def sstable_marking_test(self):
cluster = self.cluster
cluster.populate(3).start()
node1, node2, node3 = cluster.nodelist()
node3.stop(gently=True)
node1.stress(['write', 'n=10000', '-schema', 'replication(factor=3)'])
node1.flush()
node2.flush()
node3.start(wait_other_notice=True)
time.sleep(3)
if cluster.version() >= "2.2":
node3.repair()
else:
node3.nodetool("repair -par -inc")
output = ""
with open('sstables.txt', 'w') as f:
node1.run_sstablemetadata(output_file=f, keyspace='keyspace1')
node2.run_sstablemetadata(output_file=f, keyspace='keyspace1')
node3.run_sstablemetadata(output_file=f, keyspace='keyspace1')
with open("sstables.txt", 'r') as r:
output = r.read().replace('\n', '')
self.assertNotIn('Repaired at: 0', output)
os.remove('sstables.txt')
def multiple_repair_test(self):
cluster = self.cluster
cluster.populate(3).start()
node1, node2, node3 = cluster.nodelist()
session = self.patient_cql_connection(node1)
self.create_ks(session, 'ks', 3)
self.create_cf(session, 'cf', read_repair=0.0, columns={'c1': 'text', 'c2': 'text'})
debug("insert data")
insert_c1c2(session, keys=range(1, 50), consistency=ConsistencyLevel.ALL)
node1.flush()
debug("bringing down node 3")
node3.flush()
node3.stop(gently=False)
debug("inserting additional data into node 1 and 2")
insert_c1c2(session, keys=range(50, 100), consistency=ConsistencyLevel.TWO)
node1.flush()
node2.flush()
debug("restarting and repairing node 3")
node3.start(wait_for_binary_proto=True)
if cluster.version() >= "2.2":
node3.repair()
else:
node3.nodetool("repair -par -inc")
debug("stopping node 2")
node2.stop(gently=False)
debug("inserting data in nodes 1 and 3")
insert_c1c2(session, keys=range(100, 150), consistency=ConsistencyLevel.TWO)
node1.flush()
node3.flush()
debug("start and repair node 2")
node2.start(wait_for_binary_proto=True)
if cluster.version() >= "2.2":
node2.repair()
else:
node2.nodetool("repair -par -inc")
debug("replace node and check data integrity")
node3.stop(gently=False)
node5 = Node('node5', cluster, True, ('127.0.0.5', 9160), ('127.0.0.5', 7000), '7500', '0', None, ('127.0.0.5', 9042))
cluster.add(node5, False)
node5.start(replace_address='127.0.0.3', wait_other_notice=True)
assert_one(session, "SELECT COUNT(*) FROM ks.cf LIMIT 200", [149])
def sstable_repairedset_test(self):
cluster = self.cluster
cluster.populate(2).start()
node1, node2 = cluster.nodelist()
node1.stress(['write', 'n=10000', '-schema', 'replication(factor=2)'])
node1.flush()
node2.flush()
node2.stop(gently=False)
node2.run_sstablerepairedset(keyspace='keyspace1')
node2.start(wait_for_binary_proto=True)
with open('initial.txt', 'w') as f:
node2.run_sstablemetadata(output_file=f, keyspace='keyspace1')
node1.run_sstablemetadata(output_file=f, keyspace='keyspace1')
node1.stop()
node2.stress(['write', 'n=15000', '-schema', 'replication(factor=2)'])
node2.flush()
node1.start(wait_for_binary_proto=True)
if cluster.version() >= "2.2":
node1.repair()
else:
node1.nodetool("repair -par -inc")
with open('final.txt', 'w') as h:
node1.run_sstablemetadata(output_file=h, keyspace='keyspace1')
node2.run_sstablemetadata(output_file=h, keyspace='keyspace1')
with open('final.txt', 'r') as r:
finaloutput = r.read()
matches = findall('(?<=Repaired at:).*', finaloutput)
debug(matches)
uniquematches = []
matchcount = []
for value in matches:
if value not in uniquematches:
uniquematches.append(value)
matchcount.append(1)
else:
index = uniquematches.index(value)
matchcount[index] = matchcount[index] + 1
self.assertGreaterEqual(len(uniquematches), 2)
self.assertGreaterEqual(max(matchcount), 2)
self.assertNotIn('repairedAt: 0', finaloutput)
os.remove('initial.txt')
os.remove('final.txt')
@since('2.1')
def compaction_test(self):
cluster = self.cluster
cluster.populate(3).start()
node1, node2, node3 = cluster.nodelist()
session = self.patient_cql_connection(node1)
self.create_ks(session, 'ks', 3)
session.execute("create table tab(key int PRIMARY KEY, val int);")
node3.stop()
for x in range(0, 100):
session.execute("insert into tab(key,val) values(" + str(x) + ",0)")
node1.flush()
node3.start(wait_for_binary_proto=True)
if cluster.version() >= "2.2":
node3.repair()
else:
node3.nodetool("repair -par -inc")
for x in range(0, 150):
session.execute("insert into tab(key,val) values(" + str(x) + ",1)")
node1.flush()
node2.flush()
node3.flush()
node3.nodetool('compact')
for x in range(0, 150):
assert_one(session, "select val from tab where key =" + str(x), [1])
@since('2.1')
@attr('long')
@skip('hangs CI')
def multiple_subsequent_repair_test(self):
"""
Covers CASSANDRA-8366
There is an issue with subsequent inc repairs.
"""
cluster = self.cluster
cluster.populate(3).start()
[node1, node2, node3] = cluster.nodelist()
debug("Inserting data with stress")
node1.stress(['write', 'n=5M', '-rate', 'threads=10', '-schema', 'replication(factor=3)'])
debug("Flushing nodes")
cluster.flush()
debug("Waiting compactions to finish")
cluster.wait_for_compactions()
if self.cluster.version() >= '2.2':
debug("Repairing node1")
node1.nodetool("repair")
debug("Repairing node2")
node2.nodetool("repair")
debug("Repairing node3")
node3.nodetool("repair")
else:
debug("Repairing node1")
node1.nodetool("repair -par -inc")
debug("Repairing node2")
node2.nodetool("repair -par -inc")
debug("Repairing node3")
node3.nodetool("repair -par -inc")
# Using "print" instead of debug() here is on purpose. The compactions
# take a long time and don't print anything by default, which can result
# in the test being timed out after 20 minutes. These print statements
# prevent it from being timed out.
print "compacting node1"
node1.compact()
print "compacting node2"
node2.compact()
print "compacting node3"
node3.compact()
# wait some time to be sure the load size is propagated between nodes
debug("Waiting for load size info to be propagated between nodes")
time.sleep(45)
load_size_in_kb = float(sum(map(lambda n: n.data_size(), [node1, node2, node3])))
load_size = load_size_in_kb / 1024 / 1024
debug("Total Load size: {}GB".format(load_size))
# There is still some overhead, but it's lot better. We tolerate 25%.
expected_load_size = 4.5 # In GB
assert_almost_equal(load_size, expected_load_size, error=0.25)