[
https://issues.apache.org/jira/browse/HADOOP-9989?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14112608#comment-14112608
]
zhihai xu commented on HADOOP-9989:
-----------------------------------
[~daryn] I think there is another way to fix this problem using
"mapreduce.job.credentials.binary" instead of "mapreduce.job.credentials.json"
for
-tokenCacheFile in the following code:
{code}
UserGroupInformation.getCurrentUser().addCredentials(
Credentials.readTokenStorageFile(p, conf));
conf.set("mapreduce.job.credentials.json", p.toString(),
"from -tokenCacheFile command line option");
{code}
Currently the -tokenCacheFile option is broken for both MapReduce MR2 and MR1:
The following code in JobSubmitter.java will parse the file from
-tokenCacheFile as JSON file not as binary file.
But Credentials.readTokenStorageFile(p, conf) will expect it as binary file not
JSON file.
{code}
String tokensFileName = conf.get("mapreduce.job.credentials.json");
if(tokensFileName != null) {
LOG.info("loading user's secret keys from " + tokensFileName);
String localFileName = new Path(tokensFileName).toUri().getPath();
boolean json_error = false;
try {
// read JSON
ObjectMapper mapper = new ObjectMapper();
Map<String, String> nm =
mapper.readValue(new File(localFileName), Map.class);
for(Map.Entry<String, String> ent: nm.entrySet()) {
credentials.addSecretKey(new Text(ent.getKey()), ent.getValue()
.getBytes(Charsets.UTF_8));
}
} catch (JsonMappingException e) {
json_error = true;
} catch (JsonParseException e) {
json_error = true;
}
if(json_error)
LOG.warn("couldn't parse Token Cache JSON file with user secret keys");
}
}
{code}
> Bug introduced in HADOOP-9374, which parses the -tokenCacheFile as binary
> file but set it to the configuration as JSON file.
> ----------------------------------------------------------------------------------------------------------------------------
>
> Key: HADOOP-9989
> URL: https://issues.apache.org/jira/browse/HADOOP-9989
> Project: Hadoop Common
> Issue Type: Bug
> Components: security, util
> Affects Versions: 2.1.0-beta
> Environment: Red Hat Enterprise 6 with Sun Java 1.7 and IBM Java 1.6
> Reporter: Jinghui Wang
> Attachments: HADOOP-9989.patch
>
> Original Estimate: 0h
> Remaining Estimate: 0h
>
> The code in JIRA HADOOP-9374's patch introduced a bug, where the value of the
> tokenCacheFile parameter is being parsed as a binary file and set it to the
> mapreduce.job.credentials.json parameter in GenericOptionsParser, which
> cannot be parsed by JobSubmitter when it gets the value.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
