[334] in Zephyr_Bugs
[shandon@cats.UCSC.EDU: Re: zephyr acl files]
daemon@ATHENA.MIT.EDU (Jonathan I. Kamens)
Fri Jan 17 15:14:10 1992
Date: Fri, 17 Jan 92 15:13:55 -0500
From: "Jonathan I. Kamens" <jik@pit-manager.MIT.EDU>
To: bug-zephyr@Athena.MIT.EDU
Cc: shandon@cats.ucsc.edu
Any idea of whether Tim's analysis of the problem is correct, and of
how to fix it? Are the contents of the acl files he mentions relevant
to the problem?
Also, Lucien, are you in the mood to let UCSC test your new zephyr
server :-)?
jik
----- Forwarded message
From: shandon@cats.UCSC.EDU
Date: Mon, 6 Jan 92 16:06:52 -0800
To: jik@pit-manager.MIT.EDU
Subject: Re: zephyr acl files
Well, the problem at hand is with sending hm_flush messages. One of the
empty acl files with our release is hm-ctl (or rather four files with that
substring as part of their titles).
The observed behaviour is thus:
If I log into am and say "zctl hm_flush", zhm gets that request and executes
a chunk of code (I've watched this happen with dbx) that sends a flush command
to the server.
If I log into am and say "kill -HUP {zhmpid}", then it executes the same
chunk of code.
In the first case, the flush happens. In the second case it does not. My
theory for this is that the second one gets sent as user root (that being
the owner of zhm) instead of as user shandon.
The reason this is important is that machines send the HUP to zhm when they
deactivate--I believe this is MIT's solution to our server-keeps-growing
problem.
I reinstalled the old zephyr server, rather than mucking about with the new
one any more, and the same server growth problem exists with it.
-Tim
----- End of forwarded message
