SCIENTIFIC-LINUX-USERS Archives

April 2012

SCIENTIFIC-LINUX-USERS@LISTSERV.FNAL.GOV

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Lamar Owen <[log in to unmask]>
Reply To:
Lamar Owen <[log in to unmask]>
Date:
Mon, 9 Apr 2012 10:40:02 -0400
Content-Type:
Text/Plain
Parts/Attachments:
Text/Plain (38 lines)
On Thursday, April 05, 2012 08:56:20 PM you wrote:
> On Thu, Apr 5, 2012 at 3:46 PM, Ken Teh <[log in to unmask]> wrote:
> 
> > Is it true that the network manager service turns off the network when
> > there is no activity?

No.  Only if not configured to be active for all users will it not come up on boot.  There may be power management settings that can turn off the networking if unused, but on our machines I've not seen it.

> It is true that NetworkManager (one word, partly capitalized) has no
> business on any production system. It doesn't handle pair bonding, and it's
> a classic example of the sort of feature filled but configuration ruining
> and underdocumented tool that Eric Raymond ranted about years ago in his
> essay, "The Luxury of Ignorance".
> 
> Its sole use is laptops, hosts that wonder from network to network and need
> mostly-but-not-quite-useless management of VPN's and wireless access.

Upstream disagrees with you for the most part, sorry.  Desktops aren't always wired, either.  (Wireless isn't only for workstations, for that matter......).  

Having said that, it is very true that the current cluster manager 'cman' conflicts with NetworkManager, and the current NetworkManager doesn't handle many production server networking tasks, so upstream left the existing mechanism in place so choice is available.  My gut feel is that, once NM gets the features needed for cman and for proper pair bonding (among other critical features) the old mechanism will be removed.  

Whether I agree with their decision or not is really irrelevant; but it's easy enough to see their direction by following current Fedora development (although there is, and has been, some divergence between Fedora development and what has actually shown up in upstream EL).  I can see the positives for NM, and I can see the current limitations, too, so I really don't care which way they go as long as I can learn to drive it in my use case.

> > I just discovered that my desktops lost connection to the authentication
> > server. So, screen locks, gdm logins, remote ssh just stopped working.
> >  Only when I logged in as  root on the console did I notice that the
> > network was disabled and when I clicked on the the panel icon, it
> > reconnected.
...
> Yeah, it's really designed to be active only when someone is logged in, not
> for stable servers or desktops.

If you click both the 'available for all users' and the 'Connect automatically' checkboxes the interface will come up at boot without login. (As far as I remember, these checkboxes are available in Anaconda during installation if you click the button to configure netoworking; and this is thoroughly documented in upstream's documentation).  I'm running several remote desktop/VDI-type servers on EL6 where this is true, and it works fine, with multiple ethernet cards.  

I've found the EL6 implementation to be stable in a remote desktop server setting with multiple NICs and all interfaces coming up properly on boot.  I don't know why so many people are having these sorts of problems; I've not seen the majority of them in the use case for these servers.

YMMV, of course.

ATOM RSS1 RSS2