In the previous post on JUNOS I gave you a brief overview of the software architecture, with a particular emphasis on modularity. In this post, we'll have a first look at maneuvering around within a JUNOS configuration file.
The JUNOS configuration file is well organized in a hierarchical structure; once you understand that structure and its various levels, it's easy to navigate the file and find exactly the parts you want to examine or change without being distracted by parts you are not interested in at the moment.
To begin, I log into the router:
| Juniper5 (ttyd0)
login: jeff --- JUNOS 8.5R1.13 built 2007-11-14 17:54:24 UTC |
The prompt shows the user name (jeff) and the name of the router (Juniper5). The router is now in operational mode, from which you can perform operational tasks such as displaying router states and databases, ping and traceroute, telnet and ssh, restarting system processes – most commands that you need for monitoring and troubleshooting the system. The JUNOS operational mode is equivalent to the IOS user EXEC mode; in fact the prompts of both is a "greater than" (>) sign.
As with IOS, the question mark is your friend; it's the primary informational tool when you can't quite remember the next statement in a command string, or just want to see what options are available to you. I can use the question mark to see all of the commands available in operational mode:
jeff@Juniper5> ? Possible completions: |
To make configuration changes, I enter configuration mode by entering the configure command:
| jeff@Juniper5> configure
Entering configuration mode [edit] |
The JUNOS configuration mode is equivalent to the IOS privileged EXEC or "enabled" mode. Notice that the prompt changed from > to #, again the same way IOS does after you enter the enable command.
Another characteristic of the JUNOS configuration mode is the [edit] banner just above the prompt. You can maneuver around within the configuration hierarchy to observe or change just the part of the file that are of interest to you; the banner indicates precisely where you are within the hierarchy. When the banner just shows [edit], you are at the top of the hierarchy. So for example, if you type show at this level with no modiiers, JUNOS displays the entire configuration.
Using the question mark with the show command, you can see the top levels of the configuration hierarchy:
[edit] jeff@Juniper5# show ? Possible completions: <[Enter]> Execute this command > access Network access configuration > accounting-options Accounting data configuration > applications Define applications by protocol characteristics + apply-groups Groups from which to inherit configuration data > chassis Chassis configuration > class-of-service Class-of-service configuration > event-options Event processing configuration > firewall Define a firewall configuration > forwarding-options Configure options to control packet forwarding > groups Configuration groups > interfaces Interface configuration > policy-options Routing policy option configuration > protocols Routing protocol configuration > routing-instances Routing instance configuration > routing-options Protocol-independent routing option configuration > security Security configuration > services Service PIC applications settings > snmp Simple Network Management Protocol configuration > system System parameters | Pipe through a command |
The labels accompanying this list explain the levels pretty well. For example, all interfaces are configured under the interfaces level, all protocols are configured under the protocols level, and all routing policies are configured under the policy-options level.
Below each of these top levels are many sub-levels. For example, under protocols you can find BGP, OSPFv2, OSPFv3, MPLS, RSVP, and so on. And under each of those will find further sub-levels specific to the protocol. Under OSPF, for example, you will find a level for the area configurations and then under each area level you will find a level for specifying the interfaces that are in that area. I'll show you an example of that in a moment.
The hierarchical structure of the configuration allows you to focus on just the part of the configuration in which you are interested. For example, suppose you want to look at just the BGP portion of the configuration. show protocols bgp shows you the entire BGP configuration:
[edit]
jeffr@Juniper5# show protocols bgp
local-address 192.168.50.10;
local-as 65503;
group CustomerX {
export XPolicy;
peer-as 65510;
neighbor 192.168.1.1 {
authentication-key "$9$-Ud2aji.5z6qm"; ## SECRET-DATA
}
neighbor 192.168.1.5 {
authentication-key "$9$JiZHmzF/t0I69ev"; ## SECRET-DATA
}
neighbor 192.168.1.10 {
authentication-key "$9$mPF/u0Icrv1Rbs"; ## SECRET-DATA
}
}
group CustomerY {
neighbor 192.168.10.20;
neighbor 192.168.10.30;
neighbor 192.168.10.40;
}
group CustomerZ {
neighbor 192.168.20.100;
}
|
You can see in this configuration that there is a local address (192.168.50.10) from which all the BGP TCP sessions originate, the local AS number is 65503, and there are three peer groups: CustomerX, CustomerY, and CustomerZ. CustomerX has an outgoing (export) routing policy I've named XPolicy; and because the AS number for that group (65510) is different from the local AS number, we know these are EBGP peers. There are three neighbors configured under this group, all of which are authenticated with individual passwords. Peer groups CustomerY, with three neighbors, and CustomerZ, with one neighbor, are IBGP because they have no separate AS numbers specified, and no authentication is configured.
Suppose, rather than the entire BGP configuration, you are interested in only the configuration of neighbor 192.168.1.5 under group Customer X. I can tell JUNOS to display just that part of the configuration file simply by being more explicit with the show command:
[edit] jeff@Juniper5# show protocols bgp group CustomerX neighbor 192.168.1.5 authentication-key "$9$JiZHmzF/t0I69ev"; ## SECRET-DATA |
Of course if you are in configuration mode you probably want to do more than just look at parts of the configuration; you want to make changes. Using the question mark alone while in configuration mode, you can see the different actions you can take:
[edit] jeff@Juniper5# ? Possible completions: <[Enter]> Execute this command activate Remove the inactive tag from a statement annotate Annotate the statement with a comment commit Commit current set of changes copy Copy a statement deactivate Add the inactive tag to a statement delete Delete a data element edit Edit a sub-element exit Exit from this level help Provide help information insert Insert a new ordered data element load Load configuration from ASCII file quit Quit from this level rename Rename a statement replace Replace character string in configuration rollback Roll back to previous committed configuration run Run an operational-mode command save Save configuration to ASCII file set Set a parameter show Show a parameter status Show users currently editing configuration top Exit to top level of configuration up Exit one level of configuration wildcard Wildcard operations |
The remainder of this post and all of the next post discuss the actions on this list.
I want to add some elements to the configuration, and the command for doing that is set. Lets' say we want to add interface fe-0/0/0 to OSPF area 5. Using set, we specify each level down through the hierarchy until we get to the point – under the area 5 level – where the interface is to be added:
[edit] jeff@Juniper5# set protocols ospf area 5 interface fe-0/0/0 |
By the way, if OSPF was not enabled or area 5 did not yet exist, the same command would enable the protocol and create area 5 in addition to adding the interface to the area.
Another approach is to first move down to the level you want to configure, using the edit command, and then make the changes you want:
[edit] jeff@Juniper5# edit protocols ospf area 5 [edit protocols ospf area 0.0.0.5] jeff@Juniper5# set interface fe-0/0/1 |
Notice here that after entering edit protocols ospf area 5, the banner over the prompt changed to show where I am in the hierarchy – I always know where I am. I then used the set command to add interface fe-0/0/1 at that level. I don't need to specify the entire hierarchy this time, because I am already at the level where I want to make a change.
These two approaches give you the flexibility to manage a configuration in the manner most convenient to you. If you are just adding a statement or two to a configuration, it is usually easier to do that from the top level, specifying the full path down to the level you want to change. If you are making extensive changes to a particular level, it is usually easier to move to that level first using the edit command, and then make the changes directly without having to specify the full path with every command.
While edit takes you to the specific level where you want to work, up moves you up one level in the hierarchy. For instance, suppose that from where we are at the OSPF area 5 level, we want to move up to the full OSPF level:
[edit protocols ospf area 0.0.0.5] jeff@Juniper5# up [edit protocols ospf]
jeff@Juniper5# show
area 0.0.0.5 {
interface fe-0/0/0.0;
interface fe-0/0/1.0;
} |
You can see that the banner above the prompt changed after issuing the up command, telling you where you are. A show command at any level shows all of the configuration statements under the level – in this case, it shows all of the OSPF config.
If you want to jump back to the top of the hierarchy from any level, you use the top command:
[edit protocols ospf] jeff@Juniper5# top [edit] jeff@Juniper5# |
In this example, you could also have gotten from the existing level to the top by issuing the up command twice: [edit protocols ospf] => up => [edit protocols] => up => [edit]. Using the edit, up, and top commands together, you can quickly move to wherever you need to be in the configuration file for efficient operations.
In addition to the set command, there are a few more important commands for changing the configuration. In the interests of space I won't demonstrate them, but I want to at least point them out:
One of the more significant operational differences between JUNOS and IOS is that unlike IOS, when you enter a new statement into the JUNOS configuration it does not immediately become active on the router. You can make as many changes as you like, inspect your changes, and make them take effect only when you explicitly tell the router to accept them.
That's the topic of the next post: Commits, rollbacks, saves, and loads.
Jeff Doyle is president of Jeff Doyle and Associates, an IP network consultancy. Jeff is the author of Routing TCP/IP, Volumes I (read an excerpt) and II and of OSPF and IS-IS: Choosing an IGP for Large-Scale Networks. He is a frequent speaker on IPv6, MPLS, and large-scale routing.
The opinions expressed in this Weblog are those of the writer and may not represent the opinions of Network World.
|
|
HI,Jeff Thanks for your
HI,Jeff
Thanks for your posts
Indeed,we can simulate a virtual JUNOS system for study.
Ha-ha,but i'm looking forward to your posts on MPSL:
"how RSVP-TE sets up an LSP ,and, how it determines what path to use for the LSP"
"how LDP works,and difference for RSVP-TE "
Thanks & BG
Jason
----------
Welcome to
http://www.netyourlife.net
Re: MPLS
Hi Jason,
Thanks for your comments, and I'll definitely get back to covering more MPLS topics eventually.
--Jeff
Thanks a lot
Dear,Jeff
Thanks a lot,
And,I had recommanded to many many of friends in my forum to read your posts,they are all looking forward to that next next mpls topic.
Ha-HA,Thanks very much!
Thanks & BG
Jason
----------
Welcome to
http://www.netyourlife.net
JUNOS rookie
Dear Jeff:
I like your words very much. It is simple and easy to understand. I am study Juniper EX Series Switch in order to conduct the hands-on course for SI engineers. This article is a good beginning for a new JUNOS user.
BTW, thanks for your last e-mail response about OSPFv2 LSA Type 8 usage in real world!
Sincerely,
--
Johnson
Awesome post. I always
Awesome post. I always wanted to know about JUNOS.
Thanks a tonne. Anticipating more in the series :)
Best Blog on Network World
As always best content on this site or any other Networking site, from Jeff. I love Jeff's blog, he makes stuff really easy to understand - whether it is Cisco Press - Routing TCP/IP Vol1, Vol2 or new series he has started here on NW about Juniper's JunOS or his recent blogs on MPLS label switching, he simply rocks.
iBGP reason to run
Hi Jeff,
Thanks for your post in your previous article regarding the iBGP question I had posed over there. Again I do apologize for asking something completely unrelated to this article. Just had a few more questions with regards to iBGP usage in a enterprise environment.
Say if there are 4 sites with WAN Routers
P1 = Primary Site
D1 = DR Site
B1 = Remote Branch Primary Site
B2 = Remote Branch Secondary Site
Assume the following connections between the Branch and the Primary and DR Site
P1 (AS 65000)------eBGP------ B1 (AS 65001)
D1 (AS 65000)------eBGP------ B2 (AS 65001)
All these devices running BGP are the WAN Routers. These WAN routers in turn connect to other switches. The WAN Routers run EIGRP with these switches within the site and between the site.
P1 ---------- EIGRP ---------- D1
B1 ---------- EIGRP ---------- B2
The P1 and D1 sites are the Data Center Sites and they will provide transit for other Remote Branch sites to reach other.
Hence based on your previous explanation it would seem that iBGP would be required within AS65000? Is that correct? Can you elaborate on why that would be the case? Won't it be possible to just run an IGP like OSPF or EIGRP and redistributed eBGP into the IGP? Since this is not the full internet routing table but a very small subnet of private networks probably than the IGP should have no problem handling these subnets I would think?
With regards to the branch sites since they are not providing transit is iBGP required in AS65001?
I realize that the benefits of running IBGP are that you have a richer feature set to manipulate the routing. In addition with proper summarization you probably do not have to redistribute your IGP into BGP but most of the times i think you would redistribute the IGP into BGP with filters and same for vice-versa. Those are the only reasons I would see to run iBGP in a enterprise environment. Can you please advise if my thinking is correct? Thx
Re: IBGP
If I recall correctly, you asked previously why you would want to run IBGP: A very general question, to which I gave a general answer. For most general best practice rules, I could come up with exceptions, corner cases, and other reasons why the generality isn't all that important for a specific case.
Sure, you can run the above network just as you describe. Better yet, why not make it all one AS and run an IGP everywhere?
It's all about making the best design decision *for a specific network*, insuring that the risks and operational complexities are minimized, and using the best tools for the problem you are trying to solve.
--Jeff
Why does no one use Static Routes?
Using another IGP would work just fine, as Jeff responded there are lot of ways to skin the cat so to speak, its about choosing the best one.
I would ask... on network with only 4 routers, that most likely do not have redundant connections, why not consider STATIC routes?
I the understanding that routing protocols are only reasonable on a network with dozens of routers or circuits, and some redunandcy or likelyhood of change. When your network is relatively static, why not just nail down the routes you want? Static routes support manual weighting, so you can very easily have backup circuits with less weight than the primary ones. Not only that but you could save a load of CPU/resource time on the routers at the remote offices by not having to run any routing protocols like BGP or OSPF.
BGP uses a lot of memory/CPU resources. Even if you are not at the limit now, by avoiding using unneeded processes on your networking equipment you can continue to use the same equipment for longer, and wont require as many updates. You could certainly avoid using BGP on the remote offices, and likely at your data center as well. If you do have more than one connection to more than one ISP, then BGP is probably a necessity on your gateway router. If both connections are to the same ISP there are other wasy ways to use both circuits without needed BGP's giant routing tables.
I think Cisco and other experts way too often push much more complex solutions than necessary. Automation only makes sense when you are going to have to re-do the task a number of times. For large networks there is NO question that routing protocols are NECESSARY to manage all the routing and device/circuit changes that will happen. But I think the simplistic design is WAY to ofter overlooked on smaller networks.
[Jeff I would love your expert opinion on this statement]
Re; Static Routes
Hi,
In defense of the previous poster I think he was using the 4-router layout just as an example for examining interactions at protocol domain edges, not as an example of an entire network.
Having said that, I'm very much in agreement with you about static routes. I love 'em. Any kind of automatic process, whether a routing protocol, a dynamic tunneling mechanism, or whatever, means giving up some control of the network. And that means unexpected things can happen. Automatic mechanisms are needed for scale, but at routing domain edges and especially at autonomous system edges I always look to use static routes first and only go to something more dynamic if absolutely necessary.
As I tend to harp endlessly about on this blog, it's all about risk mitigation.
--Jeff
Post new comment