You are viewing a plain text version of this content. The canonical link for it is here.
Posted to hdfs-issues@hadoop.apache.org by "Allen Wittenauer (JIRA)" <ji...@apache.org> on 2012/11/29 04:34:59 UTC
[jira] [Created] (HDFS-4236) Regression: HDFS-4171 puts artificial
limit on username length
Allen Wittenauer created HDFS-4236:
--------------------------------------
Summary: Regression: HDFS-4171 puts artificial limit on username length
Key: HDFS-4236
URL: https://issues.apache.org/jira/browse/HDFS-4236
Project: Hadoop HDFS
Issue Type: Bug
Affects Versions: 2.0.3-alpha
Reporter: Allen Wittenauer
HDFS-4171 made the invalid assumption that there is a common limit on user names at the UNIX level. Almost all modern systems, when running under a 64-bit kernel, use a pointer instead of a char array in the passwd struct. This makes usernames essentially unlimited.
Additionally, IIRC, the only places where HDFS and the OS interact where the username is touched is during group lookup.
This limit is artificial and should be removed. There is a very high risk that we will break users, especially service accounts used for automated processes.
--
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
[jira] [Commented] (HDFS-4236) Regression: HDFS-4171 puts
artificial limit on username length
Posted by "Alejandro Abdelnur (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HDFS-4236?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13506223#comment-13506223 ]
Alejandro Abdelnur commented on HDFS-4236:
------------------------------------------
If anything HDFS-4171 errs by being cautious, it is not a bug. if you want to relax it please verify it won't break any things in any of the commonly used Linux distros.
> Regression: HDFS-4171 puts artificial limit on username length
> --------------------------------------------------------------
>
> Key: HDFS-4236
> URL: https://issues.apache.org/jira/browse/HDFS-4236
> Project: Hadoop HDFS
> Issue Type: Bug
> Affects Versions: 2.0.3-alpha
> Reporter: Allen Wittenauer
> Labels: regression
>
> HDFS-4171 made the invalid assumption that there is a common limit on user names at the UNIX level. Almost all modern systems, when running under a 64-bit kernel, use a pointer instead of a char array in the passwd struct. This makes usernames essentially unlimited.
> Additionally, IIRC, the only places where HDFS and the OS interact where the username is touched is during group lookup.
> This limit is artificial and should be removed. There is a very high risk that we will break users, especially service accounts used for automated processes.
--
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
[jira] [Commented] (HDFS-4236) Regression: HDFS-4171 puts
artificial limit on username length
Posted by "Suresh Srinivas (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HDFS-4236?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13508376#comment-13508376 ]
Suresh Srinivas commented on HDFS-4236:
---------------------------------------
I marked this jira as a blocker for 2.0.3, to ensure it gets picked up for the release.
> Regression: HDFS-4171 puts artificial limit on username length
> --------------------------------------------------------------
>
> Key: HDFS-4236
> URL: https://issues.apache.org/jira/browse/HDFS-4236
> Project: Hadoop HDFS
> Issue Type: Bug
> Affects Versions: 2.0.3-alpha
> Reporter: Allen Wittenauer
> Labels: regression
>
> HDFS-4171 made the invalid assumption that there is a common limit on user names at the UNIX level. Almost all modern systems, when running under a 64-bit kernel, use a pointer instead of a char array in the passwd struct. This makes usernames essentially unlimited.
> Additionally, IIRC, the only places where HDFS and the OS interact where the username is touched is during group lookup.
> This limit is artificial and should be removed. There is a very high risk that we will break users, especially service accounts used for automated processes.
--
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
[jira] [Updated] (HDFS-4236) Regression: HDFS-4171 puts artificial
limit on username length
Posted by "Suresh Srinivas (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HDFS-4236?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Suresh Srinivas updated HDFS-4236:
----------------------------------
Priority: Blocker (was: Major)
> Regression: HDFS-4171 puts artificial limit on username length
> --------------------------------------------------------------
>
> Key: HDFS-4236
> URL: https://issues.apache.org/jira/browse/HDFS-4236
> Project: Hadoop HDFS
> Issue Type: Bug
> Affects Versions: 2.0.3-alpha
> Reporter: Allen Wittenauer
> Priority: Blocker
> Labels: regression
>
> HDFS-4171 made the invalid assumption that there is a common limit on user names at the UNIX level. Almost all modern systems, when running under a 64-bit kernel, use a pointer instead of a char array in the passwd struct. This makes usernames essentially unlimited.
> Additionally, IIRC, the only places where HDFS and the OS interact where the username is touched is during group lookup.
> This limit is artificial and should be removed. There is a very high risk that we will break users, especially service accounts used for automated processes.
--
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
[jira] [Commented] (HDFS-4236) Regression: HDFS-4171 puts
artificial limit on username length
Posted by "Alejandro Abdelnur (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HDFS-4236?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13507528#comment-13507528 ]
Alejandro Abdelnur commented on HDFS-4236:
------------------------------------------
Yes, makes sense. Or we could add a config option to disable the 31 check, thus enabling longer than 31 chars for OSes where this is supported, and the default would be 31 check enabled.
> Regression: HDFS-4171 puts artificial limit on username length
> --------------------------------------------------------------
>
> Key: HDFS-4236
> URL: https://issues.apache.org/jira/browse/HDFS-4236
> Project: Hadoop HDFS
> Issue Type: Bug
> Affects Versions: 2.0.3-alpha
> Reporter: Allen Wittenauer
> Labels: regression
>
> HDFS-4171 made the invalid assumption that there is a common limit on user names at the UNIX level. Almost all modern systems, when running under a 64-bit kernel, use a pointer instead of a char array in the passwd struct. This makes usernames essentially unlimited.
> Additionally, IIRC, the only places where HDFS and the OS interact where the username is touched is during group lookup.
> This limit is artificial and should be removed. There is a very high risk that we will break users, especially service accounts used for automated processes.
--
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
[jira] [Commented] (HDFS-4236) Regression: HDFS-4171 puts
artificial limit on username length
Posted by "Suresh Srinivas (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HDFS-4236?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13507758#comment-13507758 ]
Suresh Srinivas commented on HDFS-4236:
---------------------------------------
bq. Yes, makes sense. Or we could add a config option to disable the 31 check, thus enabling longer than 31 chars for OSes where this is supported, and the default would be 31 check enabled.
Adding config option in this case seems unnecessary to me. Lets have reasonable and larger username length.
BTW can you please mark both HDFS-4171 and this jira as incompatible, add release notes to indicate the same with details and move these jiras to incompatible section in CHANGES.txt?
> Regression: HDFS-4171 puts artificial limit on username length
> --------------------------------------------------------------
>
> Key: HDFS-4236
> URL: https://issues.apache.org/jira/browse/HDFS-4236
> Project: Hadoop HDFS
> Issue Type: Bug
> Affects Versions: 2.0.3-alpha
> Reporter: Allen Wittenauer
> Labels: regression
>
> HDFS-4171 made the invalid assumption that there is a common limit on user names at the UNIX level. Almost all modern systems, when running under a 64-bit kernel, use a pointer instead of a char array in the passwd struct. This makes usernames essentially unlimited.
> Additionally, IIRC, the only places where HDFS and the OS interact where the username is touched is during group lookup.
> This limit is artificial and should be removed. There is a very high risk that we will break users, especially service accounts used for automated processes.
--
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
[jira] [Assigned] (HDFS-4236) Regression: HDFS-4171 puts artificial
limit on username length
Posted by "Allen Wittenauer (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HDFS-4236?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Allen Wittenauer reassigned HDFS-4236:
--------------------------------------
Assignee: Alejandro Abdelnur
Assigning this to Alejandro to correct.
> Regression: HDFS-4171 puts artificial limit on username length
> --------------------------------------------------------------
>
> Key: HDFS-4236
> URL: https://issues.apache.org/jira/browse/HDFS-4236
> Project: Hadoop HDFS
> Issue Type: Bug
> Affects Versions: 2.0.3-alpha
> Reporter: Allen Wittenauer
> Assignee: Alejandro Abdelnur
> Labels: regression
>
> HDFS-4171 made the invalid assumption that there is a common limit on user names at the UNIX level. Almost all modern systems, when running under a 64-bit kernel, use a pointer instead of a char array in the passwd struct. This makes usernames essentially unlimited.
> Additionally, IIRC, the only places where HDFS and the OS interact where the username is touched is during group lookup.
> This limit is artificial and should be removed. There is a very high risk that we will break users, especially service accounts used for automated processes.
--
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
[jira] [Commented] (HDFS-4236) Regression: HDFS-4171 puts
artificial limit on username length
Posted by "Allen Wittenauer (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HDFS-4236?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13506222#comment-13506222 ]
Allen Wittenauer commented on HDFS-4236:
----------------------------------------
I suppose another choice is to -1 the previous patch and revert it.
> Regression: HDFS-4171 puts artificial limit on username length
> --------------------------------------------------------------
>
> Key: HDFS-4236
> URL: https://issues.apache.org/jira/browse/HDFS-4236
> Project: Hadoop HDFS
> Issue Type: Bug
> Affects Versions: 2.0.3-alpha
> Reporter: Allen Wittenauer
> Labels: regression
>
> HDFS-4171 made the invalid assumption that there is a common limit on user names at the UNIX level. Almost all modern systems, when running under a 64-bit kernel, use a pointer instead of a char array in the passwd struct. This makes usernames essentially unlimited.
> Additionally, IIRC, the only places where HDFS and the OS interact where the username is touched is during group lookup.
> This limit is artificial and should be removed. There is a very high risk that we will break users, especially service accounts used for automated processes.
--
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
[jira] [Commented] (HDFS-4236) Regression: HDFS-4171 puts
artificial limit on username length
Posted by "Allen Wittenauer (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HDFS-4236?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13506229#comment-13506229 ]
Allen Wittenauer commented on HDFS-4236:
----------------------------------------
Alejandro, it breaks existing usernames and principals longer than 31 characters. That makes it a bug. It is a regression to previous behavior. So at a minimum requires a release note.
> Regression: HDFS-4171 puts artificial limit on username length
> --------------------------------------------------------------
>
> Key: HDFS-4236
> URL: https://issues.apache.org/jira/browse/HDFS-4236
> Project: Hadoop HDFS
> Issue Type: Bug
> Affects Versions: 2.0.3-alpha
> Reporter: Allen Wittenauer
> Labels: regression
>
> HDFS-4171 made the invalid assumption that there is a common limit on user names at the UNIX level. Almost all modern systems, when running under a 64-bit kernel, use a pointer instead of a char array in the passwd struct. This makes usernames essentially unlimited.
> Additionally, IIRC, the only places where HDFS and the OS interact where the username is touched is during group lookup.
> This limit is artificial and should be removed. There is a very high risk that we will break users, especially service accounts used for automated processes.
--
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
[jira] [Assigned] (HDFS-4236) Regression: HDFS-4171 puts artificial
limit on username length
Posted by "Alejandro Abdelnur (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HDFS-4236?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Alejandro Abdelnur reassigned HDFS-4236:
----------------------------------------
Assignee: (was: Alejandro Abdelnur)
Allen, I'd appreciate if you don't assign JIRAs to me unless I agree to. If you have special interest in fixing this please do the due diligence, verifying it will not break the various versions of Linux Hadoop is normally used and once that is done please provide a patch.
> Regression: HDFS-4171 puts artificial limit on username length
> --------------------------------------------------------------
>
> Key: HDFS-4236
> URL: https://issues.apache.org/jira/browse/HDFS-4236
> Project: Hadoop HDFS
> Issue Type: Bug
> Affects Versions: 2.0.3-alpha
> Reporter: Allen Wittenauer
> Labels: regression
>
> HDFS-4171 made the invalid assumption that there is a common limit on user names at the UNIX level. Almost all modern systems, when running under a 64-bit kernel, use a pointer instead of a char array in the passwd struct. This makes usernames essentially unlimited.
> Additionally, IIRC, the only places where HDFS and the OS interact where the username is touched is during group lookup.
> This limit is artificial and should be removed. There is a very high risk that we will break users, especially service accounts used for automated processes.
--
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