Details
-
Bug
-
Status: Open
-
Low
-
Resolution: Unresolved
-
None
-
os: ubuntu server 14.04
cassandra: 2.2.5 (apt-get installed from datastax repo)
server: ec2 m4.2xlarge
-
Low
Description
we have a cassandra cluster, which has 16 nodes(m4.2xlarge) on aws.
The total data size is 2.5TB now. There is a very large table named inbox_musical, the schema is like this:
CREATE TABLE feeds.inbox_musical (
owner_id bigint,
activity_id bigint,
insert_time timestamp,
PRIMARY KEY (owner_id, activity_id)
) WITH CLUSTERING ORDER BY (activity_id DESC)
AND bloom_filter_fp_chance = 0.1
AND caching = '{"keys":"ALL", "rows_per_partition":"20"}'
AND comment = ''
AND compaction = {'class': 'org.apache.cassandra.db.compaction.LeveledCompactionStrategy'}
AND compression = {'sstable_compression': 'org.apache.cassandra.io.compress.LZ4Compressor'}
AND dclocal_read_repair_chance = 0.1
AND default_time_to_live = 15552000
AND gc_grace_seconds = 864000
AND max_index_interval = 2048
AND memtable_flush_period_in_ms = 0
AND min_index_interval = 128
AND read_repair_chance = 0.0
AND speculative_retry = '99.0PERCENTILE';
owner_id is the partition key, and activity_id is clustering key.
95% query can return within 10ms. but a few queries read timeout ( > 5s ).
the slow queries is like this:
SELECT activity_id FROM inbox_musical WHERE owner_id=5582322 ORDER BY activity_id DESC LIMIT 100;
there are only 230 rows in this partition. but the query takes seconds to complete.
the query tracing:
activity | timestamp | source | source_elapsed
---------------------------------------------------------------------------------------------------------------------------------+----------------------------+--------------+----------------
Execute CQL3 query | 2016-05-17 11:21:26.986000 | 172.31.8.188 | 0
READ message received from /172.31.8.188 [MessagingService-Incoming-/172.31.8.188] | 2016-05-17 11:21:26.985000 | 172.31.15.14 | 13
Executing single-partition query on inbox_musical [SharedPool-Worker-4] | 2016-05-17 11:21:26.985000 | 172.31.15.14 | 115
READ message received from /172.31.8.188 [MessagingService-Incoming-/172.31.8.188] | 2016-05-17 11:21:26.986000 | 172.31.4.135 | 13
Acquiring sstable references [SharedPool-Worker-4] | 2016-05-17 11:21:26.986000 | 172.31.15.14 | 137
Executing single-partition query on inbox_musical [SharedPool-Worker-1] | 2016-05-17 11:21:26.986000 | 172.31.4.135 | 113
Merging memtable tombstones [SharedPool-Worker-4] | 2016-05-17 11:21:26.986000 | 172.31.15.14 | 161
Parsing SELECT activity_id FROM inbox_musical WHERE owner_id=5582322 ORDER BY activity_id DESC LIMIT 100; [SharedPool-Worker-3] | 2016-05-17 11:21:26.986000 | 172.31.8.188 | 66
Acquiring sstable references [SharedPool-Worker-1] | 2016-05-17 11:21:26.986000 | 172.31.4.135 | 136
Key cache hit for sstable 18948 [SharedPool-Worker-4] | 2016-05-17 11:21:26.986000 | 172.31.15.14 | 199
Preparing statement [SharedPool-Worker-3] | 2016-05-17 11:21:26.986000 | 172.31.8.188 | 99
Merging memtable tombstones [SharedPool-Worker-1] | 2016-05-17 11:21:26.986000 | 172.31.4.135 | 155
Seeking to partition beginning in data file [SharedPool-Worker-4] | 2016-05-17 11:21:26.986000 | 172.31.15.14 | 205
Read-repair DC_LOCAL [SharedPool-Worker-3] | 2016-05-17 11:21:26.986000 | 172.31.8.188 | 142
Bloom filter allows skipping sstable 20159 [SharedPool-Worker-1] | 2016-05-17 11:21:26.986000 | 172.31.4.135 | 179
Key cache hit for sstable 18924 [SharedPool-Worker-4] | 2016-05-17 11:21:26.986000 | 172.31.15.14 | 459
reading data from /172.31.4.135 [SharedPool-Worker-3] | 2016-05-17 11:21:26.986000 | 172.31.8.188 | 160
Key cache hit for sstable 20134 [SharedPool-Worker-1] | 2016-05-17 11:21:26.986000 | 172.31.4.135 | 195
Seeking to partition beginning in data file [SharedPool-Worker-4] | 2016-05-17 11:21:26.986000 | 172.31.15.14 | 468
reading data from /172.31.15.14 [SharedPool-Worker-3] | 2016-05-17 11:21:26.986000 | 172.31.8.188 | 212
Seeking to partition beginning in data file [SharedPool-Worker-1] | 2016-05-17 11:21:26.986001 | 172.31.4.135 | 202
Key cache hit for sstable 18547 [SharedPool-Worker-4] | 2016-05-17 11:21:26.986001 | 172.31.15.14 | 782
Sending READ message to /172.31.4.135 [MessagingService-Outgoing-/172.31.4.135] | 2016-05-17 11:21:26.986001 | 172.31.8.188 | 249
Key cache hit for sstable 19782 [SharedPool-Worker-1] | 2016-05-17 11:21:26.986001 | 172.31.4.135 | 376
Seeking to partition beginning in data file [SharedPool-Worker-4] | 2016-05-17 11:21:26.986001 | 172.31.15.14 | 792
Sending READ message to /172.31.15.14 [MessagingService-Outgoing-/172.31.15.14] | 2016-05-17 11:21:26.986001 | 172.31.8.188 | 260
Seeking to partition beginning in data file [SharedPool-Worker-1] | 2016-05-17 11:21:26.986001 | 172.31.4.135 | 388
Skipped 0/3 non-slice-intersecting sstables, included 0 due to tombstones [SharedPool-Worker-4] | 2016-05-17 11:21:26.986001 | 172.31.15.14 | 963
Key cache hit for sstable 16500 [SharedPool-Worker-1] | 2016-05-17 11:21:26.986001 | 172.31.4.135 | 477
Merging data from memtables and 3 sstables [SharedPool-Worker-4] | 2016-05-17 11:21:26.986001 | 172.31.15.14 | 973
Seeking to partition beginning in data file [SharedPool-Worker-1] | 2016-05-17 11:21:26.986001 | 172.31.4.135 | 484
Skipped 0/4 non-slice-intersecting sstables, included 0 due to tombstones [SharedPool-Worker-1] | 2016-05-17 11:21:26.986001 | 172.31.4.135 | 573
Merging data from memtables and 3 sstables [SharedPool-Worker-1] | 2016-05-17 11:21:26.986001 | 172.31.4.135 | 582
Read 101 live and 0 tombstone cells [SharedPool-Worker-1] | 2016-05-17 11:21:31.805000 | 172.31.4.135 | 819660
Enqueuing response to /172.31.8.188 [SharedPool-Worker-1] | 2016-05-17 11:21:31.831000 | 172.31.4.135 | 844972
Sending REQUEST_RESPONSE message to /172.31.8.188 [MessagingService-Outgoing-/172.31.8.188] | 2016-05-17 11:21:31.868000 | 172.31.4.135 | 882380
REQUEST_RESPONSE message received from /172.31.4.135 [MessagingService-Incoming-/172.31.4.135] | 2016-05-17 11:21:31.962000 | 172.31.8.188 | 976119
Processing response from /172.31.4.135 [SharedPool-Worker-2] | 2016-05-17 11:21:31.962000 | 172.31.8.188 | 976166
Request c
omplete | 2016-05-17 11:21:32.014093 | 172.31.8.188 | 28093
I used jstack to get some stacks when c* is execute query like above. and found it seems c* take a long time do the following methods:
"SharedPool-Worker-5" #221 daemon prio=5 os_prio=0 tid=0x00007f177517a000 nid=0x8b6 runnable [0x00007f1471e5b000] java.lang.Thread.State: RUNNABLE at org.apache.cassandra.db.RangeTombstoneList.addInternal(RangeTombstoneList.java:689) at org.apache.cassandra.db.RangeTombstoneList.insertFrom(RangeTombstoneList.java:649) at org.apache.cassandra.db.RangeTombstoneList.add(RangeTombstoneList.java:170) at org.apache.cassandra.db.RangeTombstoneList.add(RangeTombstoneList.java:142) at org.apache.cassandra.db.DeletionInfo.add(DeletionInfo.java:239) at org.apache.cassandra.db.ArrayBackedSortedColumns.delete(ArrayBackedSortedColumns.java:490) at org.apache.cassandra.db.ColumnFamily.addAtom(ColumnFamily.java:153) at org.apache.cassandra.db.filter.QueryFilter$2.getNext(QueryFilter.java:185) at org.apache.cassandra.db.filter.QueryFilter$2.hasNext(QueryFilter.java:157) at org.apache.cassandra.utils.MergeIterator$Candidate.advance(MergeIterator.java:146) at org.apache.cassandra.utils.MergeIterator$ManyToOne.advance(MergeIterator.java:125) at org.apache.cassandra.utils.MergeIterator$ManyToOne.computeNext(MergeIterator.java:99) at com.google.common.collect.AbstractIterator.tryToComputeNext(AbstractIterator.java:143) at com.google.common.collect.AbstractIterator.hasNext(AbstractIterator.java:138) at org.apache.cassandra.db.filter.SliceQueryFilter.collectReducedColumns(SliceQueryFilter.java:264) at org.apache.cassandra.db.filter.QueryFilter.collateColumns(QueryFilter.java:109) at org.apache.cassandra.db.filter.QueryFilter.collateOnDiskAtom(QueryFilter.java:82) at org.apache.cassandra.db.filter.QueryFilter.collateOnDiskAtom(QueryFilter.java:69) at org.apache.cassandra.db.CollationController.collectAllData(CollationController.java:317) at org.apache.cassandra.db.CollationController.getTopLevelColumns(CollationController.java:61) at org.apache.cassandra.db.ColumnFamilyStore.getTopLevelColumns(ColumnFamilyStore.java:2011) at org.apache.cassandra.db.ColumnFamilyStore.getColumnFamily(ColumnFamilyStore.java:1815) at org.apache.cassandra.db.Keyspace.getRow(Keyspace.java:360) at org.apache.cassandra.db.SliceFromReadCommand.getRow(SliceFromReadCommand.java:85)
I executed these query when server was on very low load.
from the tracing it seems that something is very slow between 'Merging data from memtables and 3 sstables' and 'Read 101 live and 0 tombstone cells'