userdel bug?

[Maki_H_]

Hi,

In MPSS 1-2.1.4982-15-rhel-6.3,I added 2 users named "hpc" and "hpc2" using micctrl as:

/usr/sbin/micctrl --useradd=hpc mic0
/usr/sbin/micctrl --useradd=hpc2 mic0

Then I deleted only "hpc" as:

/usr/sbin/micctrl --userdel=hpc mic0

After this, I got 2 lines

mic0: User hpc removed
mic0: User hpc removed

and both "hpc" and "hpc2" are removed!

[Frances_R_Intel]

It should not have let you add the user without specifying a user id (--uid=<uid>). Are there hpc and hpc2 entries in your host /etc/passwd files? Can you try adding those users to mic0 again and tell me what uid numbers were assigned to them? (The passwd file for mic0 can be found in /opt/intel/mic/filesystem/mic0/etc/passwd on the host.)  I'm just wondering if the uids matched, in which case, as far as the system was concerned, they were the same user.

[Henry_J_]

thank for you share

[Maki_H_]

Thanks for reply. I'll try and check again.

[Michael_R_]

It looks like this bug is still around and is potentially affecting us. We actually see it with --useradd. We have some local user accounts in the host passwd file that are appened with _l , but most real user accounts are in LDAP. Since there doesn't appear to be a way for the MIC cards to use LDAP directly yet I'm adding them in an init script. The problem is that if a user has a _l account already imported from the hosts passwd file the useradd does work.

For example:

micctrl --useradd=test_l --uid=600 --gid=600 mic0

micctrl --useradd=test --uid=601 --gid=601 mic0
mic0: User test already in passwd file /opt/intel/mic/filesystem/mic0/etc/passwd

It looks kind of like a regex isn't getting anchored at the end which is causing usersnames with the same prefix to fail.

[Michael_R_]

Is this bug ever going to get fixed? We just upgraded to MPSS 3.1.2 and the useradd and userdel options still don't properly select users with similar usernames. mrobbert_l and mrobbert are treated as essentially the same user to those options. 

Thanks,

Mike

[TaylorIoTKidd]

Hi Mike,

Please excuse the delay; I wanted to get you something definitive.

The fix is in the next 3.1 update coming very soon. And it is in the 3.2 release scheduled for late Q1.

Regards
--
Taylor