Re: webmin-0.990-1 spawns several processes and locks server

[Jamie C. <jca...@we...>]

Interesting .. is there any different between the working and
non-working servers, such as a different distribution or kernel version?

  - Jamie

Andrew Kornak wrote:

> Jamie,
> 
> I am connecting to a server from a remote location on the internet.
> I have a broadband internet service provider and the server is on
> a T1. No delays between myself and any servers I connect to.
> I have other servers with webmin running on the same network
> without any problems.
> 
> -Andrew
> ----- Original Message -----
> From: "Jamie Cameron" <jca...@we...>
> To: <web...@li...>
> Sent: Tuesday, September 10, 2002 11:22 PM
> Subject: Re: webmin-0.990-1 spawns several processes and locks server
> 
> 
> 
>>Joe Cooper wrote:
>>
>>
>>>Goodness!  That's not the same issue I'm seeing at all.  That's crazy
>>>talk, Andrew!
>>>
>>>I think you must have a very badly behaved module on your hands.  How
>>>about doing a Webmin session without touching any non-standard modules
>>>and seeing what happens.  I can't imagine any of the standard modules
>>>would behave that way.
>>>
>>
>>I have heard from others with this problem, and they are just using
>>
>>the standard webmin with no extra modules.. It seems to be a problem with
>>
>>miniserv.pl , not with the module code. The annoying thing is that
>>
>>I have been able to connect to the machines of people who had the problem,
>>
>>without triggering it myself!
>>
>>That leads me to suspect that it is browser or maybe network-latency
>>
> related.
> 
>>Andrew - what kind of connection is there between your browser and the
>>
> webmin
> 
>>server? Are they on the same machine, connected via a LAN, on over the
>>internet?
>>
>>
>>
>>>The problem I've seen is pretty controllable--and only leads to big
>>>trouble with one of my own modules (my troublemaker accesses a number of
>>>remote servers and gathers information from them, and then reloads after
>>>30 seconds and does it all over again infinitely many times).  I have
>>>never seen a Webmin go evil like that before.
>>>
>>
>>In your case, you might want to try disabling keep-alives in webmin, which
>>will return it to the one-process-per-request behaviour. You can do this
>>by adding the line nokeepalive=1 to /etc/webmin/miniserv.conf and then
>>re-starting.
>>
>>I'm going to look into a patch to have miniserv.pl proceses exit after 10
>>minutes, even if the client browser is till connected. It's tricky though,
>>as IE can get annoyed sometimes if the server just disconnects like that.
>>
>>  - Jamie
>>
>>
>>
>>
>>>Andrew Kornak wrote:
>>>
>>>
>>>>This might be a good short-term fix, except, my server spawns several
>>>>webmin processes in short order which brings the server down in about
>>>>5-10 seconds (not minutes). Initially, I thought perhaps I was short of
>>>>memory. I added memory and webmin sucked it right up. Not a memory
>>>>problem.
>>>>
>>>>-Andrew
>>>>----- Original Message -----
>>>>From: "Joe Cooper" <jo...@sw...>
>>>>To: <web...@li...>
>>>>Sent: Tuesday, September 10, 2002 9:41 PM
>>>>Subject: Re: webmin-0.990-1 spawns several processes and locks server
>>>>
>>>>
>>>>
>>>>
>>>>>This one is a weird issue, and one that I've seen quite a few times
>>>>>
> now,
> 
>>>>>but haven't reproduced it lately myself (but I still have a client
>>>>>
> that
> 
>>>>>can trigger it on his machine--while I cannot).
>>>>>
>>>>>I was able to make it happen in the past with Netscape 6.0, but these
>>>>>days, neither Mozilla or current Netscape makes it happen for me.
>>>>>Another client /always/ has about 25 Webmin processes on his machine,
>>>>>and I don't know where they're coming from.  Darnedest thing I've seen
>>>>>in a while...
>>>>>
>>>>>Jamie, would it be possible to get a forced timeout on child
>>>>>
> miniserv.pl
> 
>>>>>processes if they don't do something for some number of minutes?  It
>>>>>
> is
> 
>>>>>just a simple 'process per connection' server, isn't it?  I know that
>>>>>webmin spawning processes that don't end can cause it to hang around,
>>>>>but if after 5 minutes something hasn't returned, the browser has
>>>>>probably timed out the request already anyway.  So no use waiting for
>>>>>
> a
> 
>>>>>long process if there will be no one around to see it finish.  Just a
>>>>>thought.