[3249] in Release_Engineering
aklog performance & vacation & release comments
daemon@ATHENA.MIT.EDU (Richard Basch)
Tue Jun 21 16:46:12 1994
Date: Tue, 21 Jun 1994 16:46:08 -0400
To: rel-eng@MIT.EDU
From: "Richard Basch" <basch@MIT.EDU>
I do not see any slow-down in aklog for AIX.
For Solaris 2.3, I did notice a slowdown, and that has been compensated
for by setting up the "service" volume in the athena cell to contain the
OLD aklog binary. Basically, I am blessing the old binary for use with
Solaris 2.3, since the interfaces have not changed, and there are no
functional differences. I will investigate the problem further with
respect to other interfaces that are in afsuser, but since these are
used less regularly, I am not as concerned about those as I am about
other problems affecting many users.
I will be on vacation over the next few weeks.
If you need to reach me:
6/22-6/26 urgent voice mail
6/27 email accessible
7/2-7/6 1-904-343-1361
about 7/11 I will be back.
The above is certain. I may periodically check my email at other times,
probably around 7/9-7/10. The above outlines a way that I can be
reached somewhat, where my unreachability is less than one week.
Other things to note:
IBM will be shipping the updated xlc compiler to me this evening (6/21)
and it should arrive 6/22. I will install it 6/26 in the rios locker (I
will be dropping by for a couple hours).
I do not know what the problem is with the lack of obtaining tokens upon
login on Solaris and AIX. There is insufficient information, at the
moment, and I have been unable to duplicate the problem with AIX. With
Solaris, I suspect the problem is that the setgroups() system call may
not have been replaced with the AFS version, thus allowing xlogin to
trash the PAG that was setup by setpag(). The directory was most likely
attached successfully with tokens, and then made inaccessible when the
user was put back in the default PAG. The solution to this would be to
do a setpag() after the setgroups(); however, this also raises the
question as to why "su" didn't do the setgroups() and trash the PAG
during my testing. These are just some things to consider; I don't
consider the Solaris issue to be a major one.
I did note that the IBM version of xlogin had not been patched as
requested to allow a maximum of 16 groups (actually 13 with setpag), so
that NFS mounts from other servers will work.
-Richard
