[CLUE-Tech] apache and samba

[Chris Tubutis]

I've experienced NFS failures in the mid 1990s and thought it was junk.
Unreliable and difficult to manage. I later changed my mind, though...
first, the network segments my machines were on were rebuilt and cleaned
up. NFS seems to be sensitive to unhealthy networks and expects certain
services - especially hostname resolution services - to always be
working perfectly. I ultimately put all my machines in the /etc/hosts
file of each machine I ran, eliminating DNS from the mixture. Maybe did
a few other tweaks, too, I don't remember at the moment. The result was,
NFS functioned perfectly for the next 4 years and I couldn't have been
happier with it. Given that Mike already has MS boxes providing services
he relies on, I hardly think he can hurt himself any more with NFS.

In general, I'm a proponent of the simplest solution that satisfies all
the needs. The more devices and protocols and daemons that get involved,
the greater the chance that something will fail. That's my experience,
YMMV.

ct

On 23 Oct, Sterling, Willard wrote:
> It is possible to run NFS on NT, I did it years ago but I forget what
> software I used now.  I think it might have been provided latter in
> Microsoft's UNIX tools for NT package.  As a general rule though I try
> to avoid Samba and NFS mounts in production they are just not reliable
> enough. If the company has the money you may want to ask for a SAN or
> EMC to use as the DFS.
> 
> -----Original Message-----
> From: Mike Staver [mailto:staver at fimble.com]
> Sent: Thursday, October 23, 2003 1:52 PM
> To: clue-tech at clue.denver.co.us
> Subject: Re: [CLUE-Tech] apache and samba
> 
> 
> Yeah, the problem with a cron job is this stuff has to be instant. 
> People upload a document, it has to immediately be accessible to 
> everyone on all 3 machines.  So, NFS might work - except for the fact 
> that the 4th box is my companies file server, which is a windows 2k
> box.
>    I have not tried Samba 3.0 yet, so maybe that is more compatible?
> 
> Chris Tubutis wrote:
> 
>> I've used the same concept in the past - a single source used by
>> multiple boxes & httpd daemons. Having those three httpd boxes out of
>> sync doesn't give one warm, fuzzy feelings. :) I've done a few things
>> in the past, none of 'em involved Samba (much). You might have your
>> CF stuff upload stuff to a 4th box, just like you're doing now. But
>> instead of having the Web servers access that via Samba, how about if
>> they access it via NFS automount? Another possibility... upload to
>> the 4th box, then have a periodic cron job that uses rdist or rsync
>> or some such to update the httpd boxes.
>> 
>> ct
>> 
>> On 23 Oct, Mike Staver wrote:
>> 
>>>I have 3 webservers set up that I have some code on that allows me to 
>>>upload documents through cold fusion for various reasons.  These 3 
>>>webservers are all mirrors of each other, and round robin dns is what 
>>>spits users to the different servers.  So, when a document is
>>>uploaded, it only gets uploaded to one of the webservers... and the
>>>mirror is no longer a mirror :)  So, my solution to this was to add
>>>a fourth machine and mount a share on it via samba and have all the
>>>webservers use that for the file repository.  I'm not sure this is
>>>the best way to do this - using samba and all.  Sometimes on the
>>>webservers when I run dmesg, I see stuff like this:
>>>
>>>smb_trans2_request: result=-104, setting invalid
>>>smb_retry: successful, new pid=5679, generation=25
>>>smb_trans2_request: result=-104, setting invalid
>>>smb_retry: successful, new pid=5679, generation=26
>>>smb_trans2_request: result=-104, setting invalid
>>>smb_retry: successful, new pid=5679, generation=27
>>>smb_trans2_request: result=-104, setting invalid
>>>smb_retry: successful, new pid=5679, generation=28
>>>smb_trans2_request: result=-104, setting invalid
>>>smb_retry: successful, new pid=5679, generation=29
>>>smb_trans2_request: result=-104, setting invalid
>>>smb_retry: successful, new pid=5679, generation=30
>>>
>>>And then ofcourse, there is sometimes a lag while it "remounts" the 
>>>drive after these errors, so the website is slowed down a bit by this. 
>>>But after it remounts, it's great - until it loses it's connection 
>>>again.  Is there something I can do about this, or is there a better
>>>way to have mulitiple webservers share documents?
>> 
>> 
>> _______________________________________________
>> CLUE-Tech mailing list
>> Post messages to: CLUE-Tech at clue.denver.co.us
>> Unsubscribe or manage your options:
> http://clue.denver.co.us/mailman/listinfo/clue-tech
>