Property Name |
Default Value |
Description |
Type |
Is Private? |
Is Unstable? |
Is Evolving? |
tez.runtime.cleanup.files.on.interrupt |
false |
Used only for internal testing. Strictly not recommended to be used elsewhere. This
parameter could be changed/dropped later. |
boolean |
true |
false |
true |
tez.runtime.combiner.class |
null |
Specifies a combiner class (primarily for Shuffle) |
string |
false |
false |
false |
tez.runtime.combine.min.spills |
3 |
|
integer |
false |
false |
false |
tez.runtime.compress |
null |
|
boolean |
false |
false |
false |
tez.runtime.compress.codec |
null |
|
string |
false |
false |
false |
tez.runtime.convert.user-payload.to.history-text |
false |
Value: Boolean
Whether to publish configuration information to History logger. Default false. |
string |
false |
false |
false |
tez.runtime.empty.partitions.info-via-events.enabled |
true |
|
boolean |
false |
false |
false |
tez.runtime.enable.final-merge.in.output |
true |
Expert level setting. Enable final merge in ordered (defaultsorter/pipelinedsorter) outputs.
Speculative execution needs to be turned off when disabling this parameter. //TODO: TEZ-2132 |
boolean |
false |
false |
false |
tez.runtime.group.comparator.class |
null |
|
string |
false |
false |
false |
tez.runtime.ifile.readahead |
true |
Configuration key to enable/disable IFile readahead. |
boolean |
false |
false |
false |
tez.runtime.ifile.readahead.bytes |
4194304 |
Configuration key to set the IFile readahead length in bytes. |
integer |
false |
false |
false |
tez.runtime.index.cache.memory.limit.bytes |
1048576 |
|
integer |
false |
false |
false |
tez.runtime.task.input.post-merge.buffer.percent |
null |
|
float |
false |
false |
false |
tez.runtime.internal.sorter.class |
null |
|
string |
false |
false |
false |
tez.runtime.io.sort.factor |
100 |
|
integer |
false |
false |
false |
tez.runtime.io.sort.mb |
100 |
|
integer |
false |
false |
false |
tez.runtime.key.class |
null |
|
string |
false |
false |
false |
tez.runtime.key.comparator.class |
null |
|
string |
false |
false |
false |
tez.runtime.key.secondary.comparator.class |
null |
|
string |
false |
false |
false |
tez.runtime.optimize.local.fetch |
true |
If the shuffle input is on the local host bypass the http fetch and access the files directly |
boolean |
false |
false |
false |
tez.runtime.optimize.shared.fetch |
false |
Share data fetched between tasks running on the same host if applicable |
boolean |
false |
false |
false |
tez.runtime.partitioner.class |
null |
Specifies a partitioner class, which is used in Tez Runtime components
like OnFileSortedOutput |
string |
false |
false |
false |
tez.runtime.pipelined-shuffle.enabled |
false |
Expert level setting. Enable pipelined shuffle in ordered outputs and in unordered
partitioned outputs. In ordered cases, it works with PipelinedSorter.
set tez.runtime.sort.threads to greater than 1 to enable pipelinedsorter.
Ensure to set tez.runtime.enable.final-merge.in.output=false.
Speculative execution needs to be turned off when using this parameter. //TODO: TEZ-2132 |
boolean |
false |
false |
false |
tez.runtime.pipelined.sorter.lazy-allocate.memory |
false |
Setting this to true would enable sorter
to auto-allocate memory on need basis in progressive fashion.
Setting to false would allocate all available memory during
initialization of sorter. In such cases,@link{#TEZ_RUNTIME_PIPELINED_SORTER_MIN_BLOCK_SIZE_IN_MB}
would be honored and memory specified in @link{#TEZ_RUNTIME_IO_SORT_MB}
would be initialized upfront. |
boolean |
false |
false |
false |
tez.runtime.pipelined.sorter.min-block.size.in.mb |
2000 |
Tries to allocate @link{#TEZ_RUNTIME_IO_SORT_MB} in chunks specified in
this parameter. |
integer |
false |
false |
false |
tez.runtime.pipelined.sorter.sort.threads |
2 |
|
integer |
false |
false |
false |
tez.runtime.merge.progress.records |
10000 |
|
integer |
true |
false |
true |
tez.runtime.report.partition.stats |
null |
Report partition statistics (e.g better scheduling in ShuffleVertexManager). TEZ-2496
This can be enabled/disabled at vertex level.
{@link org.apache.tez.runtime.library.api.TezRuntimeConfiguration.ReportPartitionStats}
defines the list of values that can be specified.
TODO TEZ-3303 Given ShuffleVertexManager doesn't consume precise stats
yet. So do not set the value to "precise" yet when ShuffleVertexManager is used. |
string |
false |
false |
false |
tez.runtime.shuffle.acceptable.host-fetch.failure.fraction |
0.2 |
|
float |
true |
false |
true |
tez.runtime.shuffle.batch.wait |
-1 |
Expert level setting. How long should @link{ShuffleManager} wait for batching
before sending the events in milliseconds. Set to -1 to not wait. |
integer |
false |
false |
false |
tez.runtime.shuffle.buffersize |
8192 |
|
integer |
false |
false |
false |
tez.runtime.shuffle.connect.timeout |
null |
|
integer |
false |
false |
false |
tez.runtime.shuffle.memory-to-memory.enable |
false |
|
boolean |
false |
false |
false |
tez.runtime.shuffle.ssl.enable |
false |
|
boolean |
false |
false |
false |
tez.runtime.shuffle.failed.check.since-last.completion |
true |
|
boolean |
true |
false |
true |
tez.runtime.shuffle.fetcher.use-shared-pool |
false |
|
boolean |
true |
false |
true |
tez.runtime.shuffle.fetch.buffer.percent |
0.9 |
|
float |
false |
false |
false |
tez.runtime.shuffle.fetch.failures.limit |
5 |
|
integer |
false |
false |
false |
tez.runtime.shuffle.fetch.max.task.output.at.once |
20 |
|
integer |
false |
false |
false |
tez.runtime.shuffle.fetch.verify-disk-checksum |
true |
Controls verification of data checksums when fetching data directly to
disk. Enabling verification allows the fetcher to detect corrupted data
and report the failure against the upstream task before the data reaches
the Processor and causes the fetching task to fail. |
boolean |
false |
false |
false |
tez.runtime.shuffle.host.penalty.time.limit |
600000 |
Specifies in milliseconds the maximum delay a penalized host can have before being retried,
defaults to 10 minutes. |
integer |
false |
false |
false |
tez.runtime.shuffle.keep-alive.enabled |
false |
|
boolean |
false |
false |
false |
tez.runtime.shuffle.keep-alive.max.connections |
20 |
|
integer |
false |
false |
false |
tez.runtime.shuffle.max.allowed.failed.fetch.fraction |
0.5 |
|
float |
true |
false |
true |
tez.runtime.shuffle.max.stall.time.fraction |
0.5 |
|
float |
true |
false |
true |
tez.runtime.shuffle.memory.limit.percent |
0.25 |
|
float |
false |
false |
false |
tez.runtime.shuffle.memory-to-memory.segments |
null |
|
integer |
false |
false |
false |
tez.runtime.shuffle.merge.percent |
0.9 |
|
float |
false |
false |
false |
tez.runtime.shuffle.min.failures.per.host |
4 |
|
integer |
true |
false |
true |
tez.runtime.shuffle.min.required.progress.fraction |
0.5 |
|
float |
true |
false |
true |
tez.runtime.shuffle.notify.readerror |
true |
|
boolean |
false |
false |
false |
tez.runtime.shuffle.parallel.copies |
20 |
|
integer |
false |
false |
false |
tez.runtime.shuffle.read.timeout |
180000 |
|
integer |
false |
false |
false |
tez.runtime.shuffle.src-attempt.abort.limit |
-1 |
|
integer |
true |
false |
true |
tez.runtime.shuffle.use.async.http |
false |
|
boolean |
false |
false |
false |
tez.runtime.sorter.class |
null |
String value.
Which sorter implementation to use.
Valid values:
- LEGACY
- PIPELINED ( default )
{@link org.apache.tez.runtime.library.conf.OrderedPartitionedKVOutputConfig.SorterImpl} |
string |
false |
false |
false |
tez.runtime.sort.spill.percent |
0.8 |
|
float |
false |
false |
false |
tez.runtime.unordered.output.buffer.size-mb |
100 |
Size of the buffer to use if not writing directly to disk. |
integer |
false |
false |
false |
tez.runtime.unordered.output.max-per-buffer.size-bytes |
null |
Maximum size for individual buffers used in the UnsortedPartitionedOutput.
This is only meant to be used by unit tests for now. |
integer |
true |
false |
false |
tez.runtime.unordered-partitioned-kvwriter.buffer-merge-percent |
0 |
Integer value. Percentage of buffer to be filled before we spill to disk. Default value is 0,
which will spill for every buffer. |
int |
false |
false |
false |
tez.runtime.value.class |
null |
|
string |
false |
false |
false |