Добавил:
Upload Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз: Предмет: Файл:

3troubleshootingjunos

.pdf
Скачиваний:
21
Добавлен:
09.06.2015
Размер:
32.13 Mб
Скачать

Not

Troubleshooting JUNOS Platforms

ConfiguringReproductionLoo backs

Interface loopbacks r quire configuration in JUNOS Software for most PICs and interface typ s. A small number of channelized DS3 and OC12 interfaces support the ability to initiate far-end alarm and control (FEAC)-based or T1 inband and FDL-based

forloopbacks using operational mode commands. Note that configuration is never necessa y to effect an external local-loopback with a loopback plug, or when relying on the telco to provide a line loopback (which appears as a remote loopback to the attached outer). The slide shows an exampl of a local-loopback configuration and the perational mode status display that confirms that the loopback is in place.

N te that when the telco provides a line loopback, no indication exists that a loopback is in place, unless the configured Layer 2 protocol has built-in loopback detection—for example, PPP. The routers used in this example are running Frame Relay with LMI-based keepalives disabled. As a result, a remote loopback goes undetected at the remote device, which is now talking to itself as indicated by the time-to-live (TTL) expiration messages shown (we cover the use of ping to test loopbacks on a forthcoming page):

Continued on next page.

Interface Troubleshooting • Chapter 5–21

Troubleshooting JUNOS Platforms

Configuring Loopbacks (contd.)

[edit interfaces so-0/1/1]

 

 

1

 

 

user@London# run ping 10.0.22.1 count

 

 

PING 10.0.22.1 (10.0.22.1): 56 data bytes

 

 

36

bytes from 10.0.22.2: Time to

live

exceeded

Dst

Vr HL TOS

Len

ID Flg

off TTL

Pro

cks

Src

4

5

00

0054

601b

0

0000

01

01

198c

10.0.22.2

10.0.22.1

 

 

 

 

 

 

 

Reproduction

--- 10.0.22.1 ping statistics ---

received, 100% packet loss

1 packets transmitted, 0 packets

[edit interfaces so-0/1/1]

user@London# run show interfaces so-0/1/1 | match loop

Link-level type: Frame-Relay, MTU: 4474, Clocking: Internal, SONET m de, Speed: OC3, Loopback: None, FCS: 16,

Not

for

 

Chapter 5–22 • Interface Troubleshooting

Troubleshooting JUNOS Platforms

 

 

Reproduction

 

 

 

 

 

 

 

Layer 2 Protocols and Loo backs

 

 

Many Layer 2 protocols make use of a keepalive mechanism that, among other things,

 

 

can detect the s nce of a loopback. Whether local or remote, the detection of a

 

 

loop condition results in a link down declaration for that interface. When the interface

 

for

 

 

is ma ked as down at the Data Link Layer, the related interface route is removed from

 

 

the outing table, which prevents ping testing for the duration of the loopback. (We

 

 

desc ibe ping testing over a loopback on subsequent pages.)

 

 

In m st cases you can work around this issue by configuring the interface with a

 

 

no-keepalives statement, but, as shown on the slide, this workaround only works

 

 

r the frame-relay, atm, and cisco-hdlc encapsulation types. Even with

Not

keepalives (LCP) disabled, PPP still detects the presence of a loopback when the

Network Control Protocol (NCP) attempts to negotiate Layer 3 parameters. The only way around this conditions is to change the interface’s encapsulation type for the duration of the loopback test.

Note that Ethernet-related technologies have no concept of a link-layer keepalive protocol, and they do not support the concept of a remote loopback. This information is only applicable to point-to-point interface types.

Interface Troubleshooting • Chapter 5–23

Troubleshooting JUNOS Platforms

Not

Equipment fromReproductionOth r V ndors

On equipment from some oth v ndors, you can test the operation of a WAN link by issuing pings to the rout r’s local IP address. The top of the slide shows this mode of operation.

forJuniper Netwo ks Equipment

JUNOS platf ms do not exhibit this behavior. A ping sent to the device’s local IP address d es not exit the interface, and as such, cannot be used to ascertain the

perati nal status of the line.

The next slide covers testing the line on JUNOS platforms.

Chapter 5–24 • Interface Troubleshooting

Troubleshooting JUNOS Platforms

 

 

Reproduction

 

 

 

 

 

 

 

Looping Line and K

ing Interface Up

 

 

In the example the slid

, we looped the line externally. This loop might be a hard

 

 

loop, a telco loop, or a r mote interface loop for the purposes of this example.

 

 

Because some Data Link Layer protocols detect the looped condition, and disable the

 

for

 

 

 

 

 

inte face as a result, you must use either ATM Adaptation Layer 5 (AAL5), Frame Relay,

 

 

Cisco HDLC encapsulation with keepalives turned off. PPP encapsulation generally

 

 

d es not work, because the looped condition prevents the NCP from completing its

 

 

initialization, thereby preventing a declaration of up for the interface.

Not

Pinging Remote IP Address

 

 

 

 

With the loop in place and the interface up, we now issue a ping to the address associated with the remote end of the line. The address is 10.0.10.2 in this example.

Error Returns When TTL Expires

If the line has a usable transmit and receive path, the packet returns to the local device as a result of the loop condition. Upon receiving this packet, the device once again sends the packet out the WAN interface a second time. The packet’s TTL field decreases during this process. This operation continues until the packet’s TTL reaches zero, or until a line error causes packet corruption and the resulting silent packet discard.

Continued on next page.

Interface Troubleshooting • Chapter 5–25

Troubleshooting JUNOS Platforms

Error Returns When TTL Expires (contd.)

Therefore, a good line should return Internet Control Message Protocol (ICMP) TTL expired messages for every packet sent, while a marginal line might return no TTL expired errors, or it might return TTL expired messages for a small subset of the packets sent. Packet size, TTL setting, and use of the rapid switch can affect your results as well.

Reproduction

Because the default TTL for locally generated pings is 255 on a JUNOS device, each

TTL expiration message indicates 255 successful transmissions and receptio

s of the

initial ping request, all at wire speed.

 

 

Setting the TTL to a lower value is useful when trying to determine marg nal ty

f a

line. That is, a TTL of 1 requires only a single transmission and recep

f he

packet, which is similar to the type of test performed by other vendors when p ng ng a local WAN interface.

Not

for

 

Chapter 5–26 • Interface Troubleshooting

Troubleshooting JUNOS Platforms

 

 

Reproduction

 

 

 

 

 

 

 

BERT Tests

quire a Loop

 

 

The pattern

iv through a loop is verified against the pattern sent. End-to-end

 

 

testing is difficult to coordinate. By changing the position of the loop downstream from

 

 

the device performing the test, you can locate the problem area easily. Common

 

for

 

 

 

 

 

points for looping the line are the telco demarcation point (also named demark), the

 

 

emote end, and the midpoint (with help from the carrier).

 

 

Y u can configure any of the following interfaces to execute a BERT test when the

 

 

inte face eceives a request to run this test: E1, E3, T1, T3, the channelized DS-3,

Not

OC-3, OC-12, STM-1, the channelized DS-3 IQ, E1, and OC-12 IQ.

BERT Parameters

 

 

You must configure the various parameters that influence a BERT test under the interface subject to testing. These options include the test duration (10 seconds is the default), the test pattern, and the error rate to include in the bit stream by including the bert-period, bert-algorithm, and bert-error-rate statements, respectively.

Continued on next page.

Interface Troubleshooting • Chapter 5–27

Troubleshooting JUNOS Platforms

Starting and Stopping the Test

 

Start and stop the BERT test with the test interface interface-name

 

bert-start and bert-stop commands. Note that you cannot run a BERT test on

 

an interface that is administratively enabled. To start a BERT test you must first

 

disable the interface with a set interfaces interface-name disable

 

statement. After the test completes, you can use a rollback 1, commit

 

command sequence to re-enable the interface, or you can remove the disable

 

statement with a delete interface interface-name disable stateme .

 

for

Reproduction

Not

 

 

 

Chapter 5–28 • Interface Troubleshooting

Not

Troubleshooting JUNOS Platforms

You can ch ck the sults of your BERT test using the show interfaces

CheckingReproductionBERT sults

extensive command. The slide shows the formatting and fields associated with the results of a BE T test. Most of the fields are self-explanatory, but a few fields could use some additional explanation.

forThe E or bit count field displays the number of erroneous echo replies eceived f om the remote end. The LOS field indicates pattern synchronization status. A w king BERT test requires that the receiver be in sync with the transmitter. In this example, pattern synchronization was lost once during the test; furthermore, the loss

synchronization lasted for 239 seconds according go the LOS seconds field. The display also shows that no bits were received, and as a result, that no errors were detected. The lack of received bits is likely the result of the lack of test pattern synchronization.

Note that for a BERT test to be meaningful you must be able to inject and detect errors. Only by purposely injecting an error—and then witnessing that the injected error is detected—can you be sure that the test results are valid.

Interface Troubleshooting • Chapter 5–29

Troubleshooting JUNOS Platforms

 

 

 

Reproduction

 

 

 

 

 

 

 

 

Media-Specific Int rface Troubleshooting

 

 

The slide highlights the topic we discuss next.

Not

for

 

 

 

 

 

 

 

Chapter 5–30 • Interface Troubleshooting