[
https://issues.apache.org/jira/browse/HADOOP-14237?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15943066#comment-15943066
]
ASF GitHub Bot commented on HADOOP-14237:
-----------------------------------------
Github user steveloughran commented on a diff in the pull request:
https://github.com/apache/hadoop/pull/207#discussion_r108144389
--- Diff:
hadoop-tools/hadoop-aws/src/main/java/org/apache/hadoop/fs/s3a/SharedInstanceProfileCredentialsProvider.java
---
@@ -58,6 +71,84 @@ public static SharedInstanceProfileCredentialsProvider
getInstance() {
return INSTANCE;
}
+ private AWSCredentials readCredentialsFromHDFS() {
+ try {
+ FileSystem fs = FileSystem.get(new Configuration());
+ BufferedReader br = new BufferedReader(new
InputStreamReader(fs.open(s3crednetialPath)));
+ String accessKey = br.readLine();
+ String secretKey = br.readLine();
+ String token = br.readLine();
+ AWSCredentials credentials;
+ if (StringUtils.isEmpty(accessKey) ||
StringUtils.isEmpty(secretKey)) {
+ // if there are no accessKey nor secretKey return null
+ return null;
+ } else if (StringUtils.isNotEmpty(token)) {
+ credentials = new BasicSessionCredentials(accessKey, secretKey,
token);
+ } else {
+ credentials = new BasicAWSCredentials(accessKey, secretKey);
+ }
+ return credentials;
+ } catch (Exception e) {
+ return null; // ignore the read errors
+ // throw new AmazonServiceException("Failed reading S3 credentials
from HDFS " + e.getStackTrace());
+ }
+ }
+
+ private void writeCredentialsToHDFS(AWSCredentials credentials) {
+ try {
+ // Simulate atomic write by creating a new s3credential file with
random string suffix and rename to s3crednetialPath
+ Path newS3crednetialPath = new Path(s3crednetialPath.toUri() +
RandomStringUtils.randomAlphanumeric(8));
+ FileSystem fs = FileSystem.get(new Configuration());
+ BufferedWriter br = new BufferedWriter(new
OutputStreamWriter(fs.create(newS3crednetialPath, true)));
+ String accessKey = credentials.getAWSAccessKeyId();
+ String secretKey = credentials.getAWSSecretKey();
+ String token = "";
+ if (credentials instanceof BasicSessionCredentials) {
+ token = ((BasicSessionCredentials) credentials).getSessionToken();
+ }
+ br.write(accessKey);
+ br.newLine();
+ br.write(secretKey);
+ br.newLine();
+ br.write(token);
+ br.newLine();
+ br.close();
+ fs.delete(s3crednetialPath, false);
+ fs.rename(newS3crednetialPath, s3crednetialPath);
+ } catch (Exception e) {
+ // ignore write errors
+ // throw new AmazonServiceException("Failed writing S3 credentials
from HDFS " + e.getStackTrace());
+ }
+ }
+
+ @Override
+ public AWSCredentials getCredentials() {
+ for (int retry = 0; retry < maxRetries; retry++) {
+ try {
+ AWSCredentials newCredentials = super.getCredentials();
+ // if this new credentials is different from HDFS write back
+ if (credentials == null ||
(!newCredentials.getAWSSecretKey().equals(credentials.getAWSSecretKey()))) {
+ credentials = newCredentials;
+ writeCredentialsToHDFS(credentials);
+ }
+ break;
+ } catch (Exception e) {
--- End diff --
I't use our normal Retry logic here, consider some sleep + jitter if it
really is caused by throttling
> S3A Support Shared Instance Profile Credentials Across All Hadoop Nodes
> -----------------------------------------------------------------------
>
> Key: HADOOP-14237
> URL: https://issues.apache.org/jira/browse/HADOOP-14237
> Project: Hadoop Common
> Issue Type: Bug
> Components: fs/s3
> Affects Versions: 2.8.0, 3.0.0-alpha1, 3.0.0-alpha2, 2.8.1
> Environment: EC2, AWS
> Reporter: Kazuyuki Tanimura
>
> When I run a large Hadoop cluster on EC2 instances with IAM Role, it fails
> getting the instance profile credentials, eventually all jobs on the cluster
> fail. Since a number of S3A clients (all mappers and reducers) try to get the
> credentials, the AWS credential endpoint starts responding 5xx and 4xx error
> codes.
> SharedInstanceProfileCredentialsProvider.java is sort of trying to solve it,
> but it still does not share the credentials with other EC2 nodes / JVM
> processes.
> This issue prevents users from creating Hadoop clusters on EC2
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: common-issues-h...@hadoop.apache.org
