On Jun 28, 2000 Jonathan Haynes <J.Haynes@???> wrote:
> We are seeing entries of the following form in our panic log
>
> 2000-06-28 11:49:36 user "xxxxx.response" for file existence test not
> found
> 2000-06-28 11:53:43 user "xxxxx.response" for file existence test not
> found
>
> [where xxxxx has replaced a valid alias on our system and I am only
> getting these for one particular address]
When I first saw this message (but didn't pay attention to who it was
from), I thought. "Wow, someone else is seeing that too". Anyway, I will
use this as a chance to add what little I know about this.
This problem is not entirely unfamiliar to me. But are you getting it for
"only one address" as you say, or for only one value of "xxxxx"?.
I found when I last looked at this (and made no progress) that it wasn't
regularly triggered by mail to the address xxxxxxx which would pick up
that response file.
> Firstly am I correct in thinking that this means that exim is trying to
> run some (unspecified) file existance test as the username
> xxxxx.response ?
That is what the error says (and I did look at the source to confirm
that).
> The only place that response appears in our tests is the following
> director which is used for autoreplies to addresses such as abuse etc.
>
> functional_autoreply:
> driver = smartuser
> domains = "lsearch;TABLES/local_domains.common:\
> lsearch;TABLES/local_domains.THISHOST"
> transport = autoresponders
> require_files = TABLES/forwardfiles/${local_part}.response
> user = exim
> unseen
>
>
> which should run as the user exim (and there is not a file
> xxxxx.response so nothing occurs) and appears to be working for anybody
> else.
Let me add that the problem occurred both before and after the
user = exim
was added to that director.
> The only file existance test that runs under the username is the
> standard .forward style test but that director should never be reached
> until the alias has been turned into a username (and it works for
> everybody else and -d9 on all the tests described below shows exactly
> what I would expect)
Worse still there is no username xxxxxx.response, and the last time this
problem showed up there was no user xxxxxxx either (there was an alias,
however).
> When I run an address test (-bt) on xxxxx it correctly gives me the
> username to which this delivers and when I run a test on xxxxx.response
> it gives me 'unknown local user' . I can't think of any other address
> to test and can't reproduce the error.
I wasn't able to reproduce the error reliably but when it does occur, it
seems like it happens once every few minutes, so say for one in ten
messages.
> Basically I can't reproduce the problem! I can't think of any other
> address to try and I can't think of any way to trap the SMTP
> transaction
> Can anybody think of anything else I can try?
What I would have done, had I gotten around to it, is try to correlate
the panic messages with events in the main log. Probably by looking at
slow times in the main log, and playing with a bit of perl.
But I had no idea.
I did post to the exim list and PH asked me to look at all instances of
require_file in the configuration. I was thinking of modifying the source
where that message is logged to get it to log more information (ID of
message, name of file it is trying to read, etc). As you know, this was
something that I never got around to.
-j
--
Jeffrey Goldberg
Note: I am moving and changing many addresses, please see
http://www.goldmark.org/jeff/contact.html
Relativism is the triumph of convention over truth, authority over justice
