Russell Alexander Spitzer created CASSANDRA-7608:
----------------------------------------------------
Summary: StressD doesn't can't create keyspaces with Write Command
Key: CASSANDRA-7608
URL: https://issues.apache.org/jira/browse/CASSANDRA-7608
Project: Cassandra
Issue Type: Bug
Components: Tools
Reporter: Russell Alexander Spitzer
Priority: Minor
It is impossible to run the default stress command via the dameon ./stress write
Because the column names are HeapByteBuffers so they get ignored during
serilization (no error is thrown) and then when the object is deserialized on
the server the settings.columns.names is null. This leads to a null pointer on
the dameon for what would have worked had it run locally.
Settings object on the Local machine
{code}
columns = {org.apache.cassandra.stress.settings.SettingsColumn@1465}
maxColumnsPerKey = 5
names = {java.util.Arrays$ArrayList@1471} size = 5
[0] = {java.nio.HeapByteBuffer@1478}"java.nio.HeapByteBuffer[pos=0 lim=2 cap=2]"
[1] = {java.nio.HeapByteBuffer@1483}"java.nio.HeapByteBuffer[pos=0 lim=2 cap=2]"
[2] = {java.nio.HeapByteBuffer@1484}"java.nio.HeapByteBuffer[pos=0 lim=2 cap=2]"
[3] = {java.nio.HeapByteBuffer@1485}"java.nio.HeapByteBuffer[pos=0 lim=2 cap=2]"
[4] = {java.nio.HeapByteBuffer@1486}"java.nio.HeapByteBuffer[pos=0 lim=2 cap=2]"
{code}
Setings object on the StressD Machine
{code}
columns = {org.apache.cassandra.stress.settings.SettingsColumn@810}
maxColumnsPerKey = 5
names = null
{code}
This leads to the null pointer in
{code}
Exception in thread "Thread-1" java.lang.NullPointerException
at
org.apache.cassandra.stress.settings.SettingsSchema.createKeySpacesThrift(SettingsSchema.java:94)
at
org.apache.cassandra.stress.settings.SettingsSchema.createKeySpaces(SettingsSchema.java:67)
at
org.apache.cassandra.stress.settings.StressSettings.maybeCreateKeyspaces(StressSettings.java:193)
at org.apache.cassandra.stress.StressAction.run(StressAction.java:59)
at java.lang.Thread.run(Thread.java:745)
{code}
Which refers to
{code}
for (int i = 0; i < settings.columns.names.size(); i++)
standardCfDef.addToColumn_metadata(new
ColumnDef(settings.columns.names.get(i), "BytesType"));
{code}
Possible solution:
Just use the settings.columns.namestr and convert them to byte buffers at this
point in the code.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
