使用 Spark 的 Cassandra 写入性能非常低

Question

我有一个 cassandra table，有大约 500+ 百万条记录（在 6 个节点中），现在我正在尝试使用 Amazon EMR 中的 spark-cassandra-connector 插入数据

Table结构

  CREATE TABLE dmp.dmp_user_profiles_latest (
        pid text PRIMARY KEY,
        xnid int,
        day_count map<text, int>,
        first_seen map<text, timestamp>,
        last_seen map<text, timestamp>,
        usage_count map<text, int>,
        city text,
        country text,
        lid set<text>,

    )WITH bloom_filter_fp_chance = 0.01
    AND caching = '{"keys":"NONE", "rows_per_partition":"ALL"}'
    AND comment = ''
    AND compaction = {'min_threshold': '4', 'class': 'org.apache.cassandra.db.compaction.LeveledCompactionStrategy', 'max_threshold': '32'}
    AND compression = {'chunk_length_kb': '256', 'sstable_compression': 'org.apache.cassandra.io.compress.LZ4Compressor'}
    AND dclocal_read_repair_chance = 0.1
    AND default_time_to_live = 0
    AND gc_grace_seconds = 172800
    AND max_index_interval = 2048
    AND memtable_flush_period_in_ms = 0
    AND min_index_interval = 128
    AND read_repair_chance = 0.1
    AND speculative_retry = '99.0PERCENTILE';
CREATE INDEX dmp_user_profiles_latest_app_day_count_idx ON dmp.dmp_user_profiles_latest (day_count);
CREATE INDEX dmp_user_profiles_latest_country_idx ON dmp.dmp_user_profiles_latest (country);

以下是我的 spark-submit 选项

--class com.mobi.vserv.driver.Query5kPids1
--conf spark.dynamicAllocation.enabled=true  
--conf spark.yarn.executor.memoryOverhead=1024    
--conf spark.yarn.driver.memoryOverhead=1024 
--executor-memory 1g
--executor-cores 2
--driver-memory 4g

但在日志中我看到写入 Cassandra 大约需要 4-5 分钟来加载 20 万（200,000）条记录（而总执行时间为 6 分钟以上）

我也在 Spark conf 中添加了以下内容

conf.set("spark.cassandra.output.batch.size.rows", "auto");
conf.set("spark.cassandra.output.concurrent.writes", "500");
conf.set("spark.cassandra.output.batch.size.bytes", "100000");
conf.set("spark.cassandra.output.throughput_mb_per_sec","1");

但仍然没有性能提升，增加 Amazon EMR 中的内核数也无济于事。

请注意，在我的 Cassandra table 中，我们没有使用任何 partitioning/clustering 列，这可能是性能如此缓慢的原因。

请注意网络速度为 30 MB PS 主键是字母数字值，例如 - a9be3eb4-751f-48ee-b593-b3f89e18622d

Cassandra.yaml

cluster_name: 'dmp Cluster'
num_tokens: 100
hinted_handoff_enabled: true
max_hint_window_in_ms: 10800000 # 3 hours
hinted_handoff_throttle_in_kb: 1024
max_hints_delivery_threads: 2
batchlog_replay_throttle_in_kb: 1024
authenticator: AllowAllAuthenticator
authorizer: AllowAllAuthorizer
permissions_validity_in_ms: 2000
partitioner: org.apache.cassandra.dht.Murmur3Partitioner
data_file_directories:
     - /data/cassandra/data
disk_failure_policy: stop
commit_failure_policy: stop

key_cache_size_in_mb:

key_cache_save_period: 14400
row_cache_size_in_mb: 0
row_cache_save_period: 0
counter_cache_size_in_mb:
counter_cache_save_period: 7200
saved_caches_directory: /data/cassandra/saved_caches
commitlog_sync: periodic
commitlog_sync_period_in_ms: 10000
seed_provider:
 - class_name: org.apache.cassandra.locator.SimpleSeedProvider
    parameters:
 - seeds: "10.142.76.97,10.182.19.301"

concurrent_reads: 256
concurrent_writes: 128
concurrent_counter_writes: 32

memtable_allocation_type: heap_buffers
memtable_flush_writers: 8
index_summary_capacity_in_mb:
index_summary_resize_interval_in_minutes: 60
trickle_fsync: false
trickle_fsync_interval_in_kb: 10240
storage_port: 7000
ssl_storage_port: 7001
listen_address: 10.142.76.97
start_rpc: true
rpc_address: 10.23.244.172
rpc_port: 9160
rpc_keepalive: true
rpc_server_type: sync
thrift_framed_transport_size_in_mb: 15
incremental_backups: false
snapshot_before_compaction: false
auto_snapshot: true
tombstone_warn_threshold: 1000
tombstone_failure_threshold: 100000
column_index_size_in_kb: 64
batch_size_warn_threshold_in_kb: 5
concurrent_compactors: 4
compaction_throughput_mb_per_sec: 64
sstable_preemptive_open_interval_in_mb: 50
read_request_timeout_in_ms: 500000

range_request_timeout_in_ms: 1000000

write_request_timeout_in_ms: 200000

counter_write_request_timeout_in_ms: 500000

cas_contention_timeout_in_ms: 100000

endpoint_snitch: Ec2Snitch

dynamic_snitch_update_interval_in_ms: 100

dynamic_snitch_reset_interval_in_ms: 600000

dynamic_snitch_badness_threshold: 0.1

request_scheduler: org.apache.cassandra.scheduler.NoScheduler

server_encryption_options:
    internode_encryption: none
    keystore: conf/.keystore
    keystore_password: cassandra
    truststore: conf/.truststore
    truststore_password: cassandra

client_encryption_options:
    enabled: false
    keystore: conf/.keystore
    keystore_password: cassandra

internode_compression: all

inter_dc_tcp_nodelay: false

Answer 1

正如评论中所说，您的问题似乎来自 day_count 上的索引。

如本 page 所示，如果您必须一直更新索引，索引将不会有效，而当您将不同的值插入 day_count 时（可能每次都).

您需要重新处理您的数据库，但由于这是您的生产环境，如果需要此索引，您不能只 DROP INDEX IF EXISTS keyspace.index_name，但您可以使用 day_count 创建一个辅助数据库作为主键，或使用 day_count 作为排序索引。

使用 Spark 的 Cassandra 写入性能非常低

Cassandra write giving very slow perfomance using Spark

apache-spark

cassandra

amazon-emr

spark-cassandra-connector