Spaces:
Build error
Build error
File size: 11,887 Bytes
3932407 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 |
import multiprocessing
import pathlib
import uuid
from time import sleep
from .fixtures import upsert_random_points, create_collection
from .utils import *
N_PEERS = 3
N_SHARDS = 3
N_REPLICA = 1
COLLECTION_NAME = "test_collection"
def update_points_in_loop(peer_url, collection_name, offset=0, throttle=False, duration=None):
start = time.time()
limit = 3
while True:
upsert_random_points(peer_url, limit, collection_name, offset=offset)
offset += limit
if throttle:
sleep(0.1)
if duration is not None and (time.time() - start) > duration:
break
def run_update_points_in_background(peer_url, collection_name, init_offset=0, throttle=False, duration=None):
p = multiprocessing.Process(target=update_points_in_loop, args=(peer_url, collection_name, init_offset, throttle, duration))
p.start()
return p
# Transfer shards from one node to another
#
# Simply does the most basic transfer: no concurrent updates during the
# transfer.
#
# Test that data on the both sides is consistent
def test_shard_snapshot_transfer(tmp_path: pathlib.Path):
assert_project_root()
# seed port to reuse the same port for the restarted nodes
peer_api_uris, peer_dirs, bootstrap_uri = start_cluster(tmp_path, N_PEERS, 20000)
create_collection(peer_api_uris[0], shard_number=N_SHARDS, replication_factor=N_REPLICA)
wait_collection_exists_and_active_on_all_peers(
collection_name=COLLECTION_NAME,
peer_api_uris=peer_api_uris
)
# Insert some initial number of points
upsert_random_points(peer_api_uris[0], 100)
transfer_collection_cluster_info = get_collection_cluster_info(peer_api_uris[0], COLLECTION_NAME)
receiver_collection_cluster_info = get_collection_cluster_info(peer_api_uris[2], COLLECTION_NAME)
from_peer_id = transfer_collection_cluster_info['peer_id']
to_peer_id = receiver_collection_cluster_info['peer_id']
shard_id = transfer_collection_cluster_info['local_shards'][0]['shard_id']
# Transfer shard from one node to another
# Move shard `shard_id` to peer `target_peer_id`
r = requests.post(
f"{peer_api_uris[0]}/collections/{COLLECTION_NAME}/cluster", json={
"replicate_shard": {
"shard_id": shard_id,
"from_peer_id": from_peer_id,
"to_peer_id": to_peer_id,
"method": "snapshot",
}
})
assert_http_ok(r)
# Wait for end of shard transfer
wait_for_collection_shard_transfers_count(peer_api_uris[0], COLLECTION_NAME, 0)
receiver_collection_cluster_info = get_collection_cluster_info(peer_api_uris[2], COLLECTION_NAME)
number_local_shards = len(receiver_collection_cluster_info['local_shards'])
assert number_local_shards == 2
# Point counts must be consistent across nodes
counts = []
for uri in peer_api_uris:
r = requests.post(
f"{uri}/collections/{COLLECTION_NAME}/points/count", json={
"exact": True
}
)
assert_http_ok(r)
counts.append(r.json()["result"]['count'])
assert counts[0] == counts[1] == counts[2]
# Transfer shards from one node to another with an API key is configured
#
# Simply does the most basic transfer: no concurrent updates during the
# transfer.
#
# Test that data on the both sides is consistent
def test_shard_snapshot_transfer_with_api_key(tmp_path: pathlib.Path):
assert_project_root()
# Configure a random API key
api_key = str(uuid.uuid4())
env={
"QDRANT__SERVICE__API_KEY": api_key,
}
headers={
"api-key": api_key,
}
# seed port to reuse the same port for the restarted nodes
peer_api_uris, peer_dirs, bootstrap_uri = start_cluster(tmp_path, N_PEERS, 20000, extra_env=env, headers=headers)
create_collection(peer_api_uris[0], shard_number=N_SHARDS, replication_factor=N_REPLICA, headers=headers)
wait_collection_exists_and_active_on_all_peers(
collection_name=COLLECTION_NAME,
peer_api_uris=peer_api_uris,
headers=headers,
)
# Insert some initial number of points
upsert_random_points(peer_api_uris[0], 100, headers=headers)
transfer_collection_cluster_info = get_collection_cluster_info(peer_api_uris[0], COLLECTION_NAME, headers=headers)
receiver_collection_cluster_info = get_collection_cluster_info(peer_api_uris[2], COLLECTION_NAME, headers=headers)
from_peer_id = transfer_collection_cluster_info['peer_id']
to_peer_id = receiver_collection_cluster_info['peer_id']
shard_id = transfer_collection_cluster_info['local_shards'][0]['shard_id']
# Transfer shard from one node to another
# Move shard `shard_id` to peer `target_peer_id`
r = requests.post(
f"{peer_api_uris[0]}/collections/{COLLECTION_NAME}/cluster", json={
"replicate_shard": {
"shard_id": shard_id,
"from_peer_id": from_peer_id,
"to_peer_id": to_peer_id,
"method": "snapshot",
}
},
headers=headers,
)
assert_http_ok(r)
# Wait for end of shard transfer
wait_for_collection_shard_transfers_count(peer_api_uris[0], COLLECTION_NAME, 0, headers=headers)
receiver_collection_cluster_info = get_collection_cluster_info(peer_api_uris[2], COLLECTION_NAME, headers=headers)
number_local_shards = len(receiver_collection_cluster_info['local_shards'])
assert number_local_shards == 2
# Point counts must be consistent across nodes
counts = []
for uri in peer_api_uris:
r = requests.post(
f"{uri}/collections/{COLLECTION_NAME}/points/count", json={
"exact": True
},
headers=headers,
)
assert_http_ok(r)
counts.append(r.json()["result"]['count'])
assert counts[0] == counts[1] == counts[2]
# Transfer shards from one node to another while applying throttled updates in parallel
#
# Updates are throttled to prevent sending updates faster than the queue proxy
# can handle. The transfer must therefore finish in 30 seconds without issues.
#
# Test that data on the both sides is consistent
def test_shard_snapshot_transfer_throttled_updates(tmp_path: pathlib.Path):
assert_project_root()
# seed port to reuse the same port for the restarted nodes
peer_api_uris, peer_dirs, bootstrap_uri = start_cluster(tmp_path, N_PEERS, 20000)
create_collection(peer_api_uris[0], shard_number=N_SHARDS, replication_factor=N_REPLICA)
wait_collection_exists_and_active_on_all_peers(
collection_name=COLLECTION_NAME,
peer_api_uris=peer_api_uris
)
# Insert some initial number of points
upsert_random_points(peer_api_uris[0], 10000)
# Start pushing points to the cluster
upload_process_1 = run_update_points_in_background(peer_api_uris[0], COLLECTION_NAME, init_offset=100, throttle=True)
upload_process_2 = run_update_points_in_background(peer_api_uris[1], COLLECTION_NAME, init_offset=10000, throttle=True)
upload_process_3 = run_update_points_in_background(peer_api_uris[2], COLLECTION_NAME, init_offset=20000, throttle=True)
transfer_collection_cluster_info = get_collection_cluster_info(peer_api_uris[0], COLLECTION_NAME)
receiver_collection_cluster_info = get_collection_cluster_info(peer_api_uris[2], COLLECTION_NAME)
from_peer_id = transfer_collection_cluster_info['peer_id']
to_peer_id = receiver_collection_cluster_info['peer_id']
shard_id = transfer_collection_cluster_info['local_shards'][0]['shard_id']
# Transfer shard from one node to another
# Move shard `shard_id` to peer `target_peer_id`
r = requests.post(
f"{peer_api_uris[0]}/collections/{COLLECTION_NAME}/cluster", json={
"replicate_shard": {
"shard_id": shard_id,
"from_peer_id": from_peer_id,
"to_peer_id": to_peer_id,
"method": "snapshot",
}
})
assert_http_ok(r)
# Wait for end of shard transfer
wait_for_collection_shard_transfers_count(peer_api_uris[0], COLLECTION_NAME, 0)
upload_process_1.kill()
upload_process_2.kill()
upload_process_3.kill()
sleep(1)
receiver_collection_cluster_info = get_collection_cluster_info(peer_api_uris[2], COLLECTION_NAME)
number_local_shards = len(receiver_collection_cluster_info['local_shards'])
assert number_local_shards == 2
# Point counts must be consistent across nodes
counts = []
for uri in peer_api_uris:
r = requests.post(
f"{uri}/collections/{COLLECTION_NAME}/points/count", json={
"exact": True
}
)
assert_http_ok(r)
counts.append(r.json()["result"]['count'])
assert counts[0] == counts[1] == counts[2]
# Transfer shards from one node to another while applying updates in parallel
#
# A fast burst of updates is sent in the first 5 seconds, the queue proxy will
# not be able to keep up with this. After that, updates are throttled. The
# transfer must still finish in 30 seconds without issues.
#
# Test that data on the both sides is consistent
def test_shard_snapshot_transfer_fast_burst(tmp_path: pathlib.Path):
assert_project_root()
# seed port to reuse the same port for the restarted nodes
peer_api_uris, peer_dirs, bootstrap_uri = start_cluster(tmp_path, N_PEERS, 20000)
create_collection(peer_api_uris[0], shard_number=N_SHARDS, replication_factor=N_REPLICA)
wait_collection_exists_and_active_on_all_peers(
collection_name=COLLECTION_NAME,
peer_api_uris=peer_api_uris
)
# Insert some initial number of points
upsert_random_points(peer_api_uris[0], 10000)
# Start pushing points to the cluster
upload_process_1 = run_update_points_in_background(peer_api_uris[0], COLLECTION_NAME, init_offset=100, duration=5)
upload_process_2 = run_update_points_in_background(peer_api_uris[1], COLLECTION_NAME, init_offset=10000, duration=5)
upload_process_3 = run_update_points_in_background(peer_api_uris[2], COLLECTION_NAME, init_offset=20000, throttle=True)
transfer_collection_cluster_info = get_collection_cluster_info(peer_api_uris[0], COLLECTION_NAME)
receiver_collection_cluster_info = get_collection_cluster_info(peer_api_uris[2], COLLECTION_NAME)
from_peer_id = transfer_collection_cluster_info['peer_id']
to_peer_id = receiver_collection_cluster_info['peer_id']
shard_id = transfer_collection_cluster_info['local_shards'][0]['shard_id']
# Transfer shard from one node to another
# Move shard `shard_id` to peer `target_peer_id`
r = requests.post(
f"{peer_api_uris[0]}/collections/{COLLECTION_NAME}/cluster", json={
"replicate_shard": {
"shard_id": shard_id,
"from_peer_id": from_peer_id,
"to_peer_id": to_peer_id,
"method": "snapshot",
}
})
assert_http_ok(r)
# Wait for end of shard transfer
wait_for_collection_shard_transfers_count(peer_api_uris[0], COLLECTION_NAME, 0)
upload_process_1.kill()
upload_process_2.kill()
upload_process_3.kill()
sleep(1)
receiver_collection_cluster_info = get_collection_cluster_info(peer_api_uris[2], COLLECTION_NAME)
number_local_shards = len(receiver_collection_cluster_info['local_shards'])
assert number_local_shards == 2
# Point counts must be consistent across nodes
counts = []
for uri in peer_api_uris:
r = requests.post(
f"{uri}/collections/{COLLECTION_NAME}/points/count", json={
"exact": True
}
)
assert_http_ok(r)
counts.append(r.json()["result"]['count'])
assert counts[0] == counts[1] == counts[2]
|