MailScanner stopping

[Kevin Miller]

Glenn Steen wrote:
> On 13/09/06, Kevin Miller <Kevin_Miller at ci.juneau.ak.us> wrote:
>> Glenn Steen wrote:
>>> On 13/09/06, Kevin Miller <Kevin_Miller at ci.juneau.ak.us> wrote:
>>>> Last week there was a thread about MailScanner ceasing to work.  I
>>>> was seeing that behavior - it was running, but was all balled up.
>>>> Turned out that when 'MailScanner restart' was called that some
>>>> sendmail processes persisted past the $RESTART_DELAY time.  This
>>>> only appears to happen on later MS servers.  Not sure what changes
>>>> to the init script make the difference - my 4.48 server doesn't
>>>> have a problem. 
>>>> 
>>>> The pids that MailScanner creates all evaporate almost immediately;
>>>> the persistent sendmail processes are generally (as nearly as I can
>>>> tell) connections that smf-sav initiates.  (smf-sav is sorta like
>>>> milter-sender and milter-ahead rolled into one.)  I wasn't seeing
>>>> the problem until I installed it, but it's too nifty to get rid of
>>>> even if it's not as robust as the milters from Snertsoft.
>>>> 
>>>> It can take up to a minute or two for all the connections to
>>>> expire. I'm guessing the remote hosts are maybe doing greet-pause
>>>> or graylisting when I do the sender verification, which causes
>>>> them to sit around for a bit. 
>>>> 
>>> Sounds a bit odd... You're not getting tar-pitted by your own
>>> internal servers, are you?
>> 
>> Nope - doing a ps aux shows the connections and associated far end.
>> Most connections close pretty fast, there's just one or two, maybe
>> three that persist.  Not the same one's of course, since at any
>> given random moment I could be connected to almost anyone.
> 
> Fair enough.
> 
>> The idea that some sendmail processes persist longer than the
>> $RESTART_DELAY has been mentioned in other contexts for a long time.
>> It's just that it was a very random occurance historically.  Now (at
>> least in my case) it is more frequent.
> 
> And I'm sure you've scrutinized every possible angle (and some
> impossible:-) as to why it's more frequent now, so I'll shut up about
> that:-)
> 
>> I assume that in general it's not a good idea to restart until all is
>> cleaned up, hence the hack.
> Good idea, yes.
>>  A real bash programmer could probably make
>> it prettier.
> Pretty doesn't come into it when scripting... Usual code health, yes,
> but not "pretty":-). After all, when scripting in <insert favourite
> shell> we're after function... If we wanted pretty we'd not do it in a
> shell:-):-)
>>  For some reason the newline doesn't kick in when I echo
>> the the output of ps.  Don't know why.  Don't really need the echos
>> in there at all, but I find them handy to see where I'm stalled,
>> even if the formatting is butt-ugly.
> 
> I think I know why, or at least some part of it... You have a rather
> "bad" error in that case statement... I'll just copy it here and
> comment in it:
>  restart)
>    $0 stop
>    # Initialize the test variable
>    Pidval=sendmail
>    ;; # <---- This line terminates the case started at the "restart)".
> # Everything below this line, until the next case is just _never
>    executed_ # Loop as long as there's a sendmail process
>    while [ "$Pidval" != "" ];
>    do
>          # Look for sendmail but ignore the grep
>        Pidval=`ps aux | grep sendmail | grep -v grep`
>        echo
>        echo $Pidval
>        sleep $RESTART_DELAY
>    done
>    $0 start
>    rc_status
> # This is where you should have have the double semi-colons (;;) ...
> 
> Further, some distros try to be overly clever about their sysV init
> scripts, so even with that corrected, you might get some rather
> non-obvious displays from those echos. But then, it really should work
> OK with that fixed.
> The "style" doesn't matter much, and the technique to do the waiting
> is pretty standard... What you could do is "enhance" it with a
> counter, so that you only iterate a set number of times (6 times would
> give you three minutes).

Thanks Glenn.  As it turns out, I just made that mistake in what I
posted here.  I tried it out on my servers and they worked as advertised
(that's how I knew the newline wasn't working right).  I probably copied
from the editor prior to a bit of cleanup.  By a stroke of luck all my
servers looked fine when I checked them just now.

Not sure I'd want a counter in there since the idea is to be absolutely
sure that everything is shut down bere restarting.

Appreciate the clarification none the less, as I was thinking they were
just comment delimeters similar to the # sign.  Learn something new
every day...

...Kevin
-- 
Kevin Miller                Registered Linux User No: 307357
CBJ MIS Dept.               Network Systems Admin., Mail Admin.
155 South Seward Street     ph: (907) 586-0242
Juneau, Alaska 99801        fax: (907 586-4500