You are viewing a plain text version of this content. The canonical link for it is here.
Posted to common-issues@hadoop.apache.org by "Suresh Srinivas (JIRA)" <ji...@apache.org> on 2013/07/11 18:57:49 UTC
[jira] [Comment Edited] (HADOOP-9688) Add globally unique Client ID
to RPC requests
[ https://issues.apache.org/jira/browse/HADOOP-9688?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13705253#comment-13705253 ]
Suresh Srinivas edited comment on HADOOP-9688 at 7/11/13 4:57 PM:
------------------------------------------------------------------
bq. storageIDs are unique. There is now probability they will collide as far as I remember. I am saying we can do the same for clientID.
The probability of StorageID collision is also non-zero. I am not sure how you can assert otherwise. If you see the randomUUID method implementation, it also uses SecureRandom class. Also unlike storageID, UUID are compact and 16 bytes.
bq. You know in big clusters most improbable events happen all the time.
I think this improbable event can also happen for StorageID.
Unlike StorageID, retry has many things that further reduce the probability:
# The clientID + callID is used only for retried requests. See HADOP-9717.
# The validity of an entry in retry cache is short (period of 10 minutes).
was (Author: sureshms):
bq. storageIDs are unique. There is now probability they will collide as far as I remember. I am saying we can do the same for clientID.
The probability of StorageID collision is also non-zero. I am not sure how you can assert otherwise. If you see the randomUUID method implementation, it also uses SecureRandom class. Also unlike storageID, UUID are compact and 16 bytes.
bq. You know in big clusters most improbable events happen all the time.
I think this improbable even can also happen for StorageID.
Unlike StorageID, retry has many things that further reduce the probability:
# The clientID + callID is used only for retried requests. See the HADOP-9717.
# The validity of an entry in retry cache is short.
> Add globally unique Client ID to RPC requests
> ---------------------------------------------
>
> Key: HADOOP-9688
> URL: https://issues.apache.org/jira/browse/HADOOP-9688
> Project: Hadoop Common
> Issue Type: Improvement
> Components: ipc
> Reporter: Suresh Srinivas
> Assignee: Suresh Srinivas
> Priority: Blocker
> Fix For: 3.0.0
>
> Attachments: HADOOP-9688.clientId.1.patch, HADOOP-9688.clientId.patch, HADOOP-9688.patch
>
>
> This is a subtask in hadoop-common related to HDFS-4942 to add unique request ID to RPC requests.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira