[
https://issues.apache.org/jira/browse/PHOENIX-2311?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14947074#comment-14947074
]
Brian Panneton commented on PHOENIX-2311:
-----------------------------------------
I'd suggest to always write to the tmp dir given by java (-Djava.io.tmpdir).
Thus you don't have to create a new flag.
> performance.py java.io.FileNotFoundException: ddl.sql
> -----------------------------------------------------
>
> Key: PHOENIX-2311
> URL: https://issues.apache.org/jira/browse/PHOENIX-2311
> Project: Phoenix
> Issue Type: Bug
> Affects Versions: 4.5.2
> Reporter: Albert Chu
> Priority: Minor
>
> Recently hit an issue where I got
> {noformat}
> java.io.FileNotFoundException: ddl.sql (No such file or directory)
> at java.io.FileInputStream.open(Native Method)
> at java.io.FileInputStream.<init>(FileInputStream.java:146)
> at java.io.FileInputStream.<init>(FileInputStream.java:101)
> at java.io.FileReader.<init>(FileReader.java:58)
> at
> org.apache.phoenix.util.PhoenixRuntime.main(PhoenixRuntime.java:222)
> {noformat}
> The issue appears to be that I have phoenix installed in a NFS home
> directory. When I run multiple performance tests simultaneously, there is a
> race on the create/delete of the temporary files in that NFS dir (such as
> ddl.sql).
> as a quick hack, I just added "/tmp/" to each of the files in performance.py
> to see if that would fix things, but then hit:
> {noformat}
> java.io.FileNotFoundException: /tmp/data.csv (No such file or directory)
> at java.io.FileInputStream.open(Native Method)
> at java.io.FileInputStream.<init>(FileInputStream.java:146)
> at org.apache.commons.csv.CSVParser.parse(CSVParser.java:158)
> at
> org.apache.phoenix.util.CSVCommonsLoader.upsert(CSVCommonsLoader.java:191)
> at
> org.apache.phoenix.util.PhoenixRuntime.main(PhoenixRuntime.java:237)
> {noformat}
> digging further, it appears in the GeneratePerformanceData class, the
> "data.csv" file is hard coded.
> {noformat}
> private static final String FILENAME = "data.csv";
> {noformat}
> I stopped digging at this point.
> There are many ways this could be fixed. Some additional optional arguments
> would probably be easiest/make sense.
> Unsure of what upstream would prefer. I'm more than happy to produce a patch
> based on upstream preferences.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
