deniskuzZ opened a new pull request, #523:
URL: https://github.com/apache/tez/pull/523
The serialized DAG plan file (tez-dag.pb<counter>) written for plans
exceeding the IPC limit is named by a per-TezClient counter and never deleted.
When the session and its staging directory outlive the TezClient instances, a
new client generation recomputes the same file name and fails with
FileAlreadyExistsException, permanently blocking oversized submissions on that
session.
Create the plan file with overwrite=true (a file at that path can only be an
already consumed leftover), and delete it in the AM once the submitDAG request
has been parsed.
Reproduced by the new unit test:
TestTezClient#testSessionLargeDAGPlanWithLeftoverPlanFile:
````
org.apache.hadoop.fs.FileAlreadyExistsException: File already exists:
file:/.../tez-api/target/tmp/org.apache.tez.client.TestTezClient/.tez/application_0_0001/tez-dag.pb1
at
org.apache.hadoop.fs.RawLocalFileSystem.create(RawLocalFileSystem.java:622)
at
org.apache.hadoop.fs.RawLocalFileSystem.create(RawLocalFileSystem.java:660)
at
org.apache.hadoop.fs.ChecksumFileSystem$ChecksumFSOutputSummer.<init>(ChecksumFileSystem.java:692)
at
org.apache.hadoop.fs.ChecksumFileSystem.create(ChecksumFileSystem.java:780)
at
org.apache.hadoop.fs.ChecksumFileSystem.create(ChecksumFileSystem.java:759)
at org.apache.hadoop.fs.FileSystem.create(FileSystem.java:1234)
at org.apache.hadoop.fs.FileSystem.create(FileSystem.java:1211)
at org.apache.hadoop.fs.FileSystem.create(FileSystem.java:1092)
at
org.apache.tez.client.TezClient.submitDAGSession(TezClient.java:695)
at org.apache.tez.client.TezClient.submitDAG(TezClient.java:614)
at
org.apache.tez.client.TestTezClient.testSessionLargeDAGPlanWithLeftoverPlanFile(TestTezClient.java:328)
````
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
