Errata

Arista Warrior

Errata for Arista Warrior

Submit your own errata for this product.

The errata list is a list of errors and their corrections that were found after the product was released.

The following errata were submitted by our customers and have not yet been approved or disproved by the author or editor. They solely represent the opinion of the customer.

Color Key: Serious technical mistake Minor technical mistake Language or formatting error Typo Question Note Update

Version Location Description Submitted by Date submitted
PDF Page xvi
4th para

I would like to especially thank Glenn Bradley with his help
I would like to especially thank Glenn Bradley for his help

Chris Pepper  Feb 27, 2013 
PDF Page xviii
p1

I?ve simply changed the point in which the line wraps
I?ve simply changed the point at which the line wraps

Chris Pepper  Feb 27, 2013 
PDF Page 1
p4

state-of-the art
state-of-the-art

Chris Pepper  Feb 27, 2013 
PDF Page 5
p4

VLANs spanning across physical locations,
VLANs spanning physical locations,

Chris Pepper  Feb 27, 2013 
Printed Page 6
4th paragraph

"NAS" is used where "NFS" should be:

wrong: The NAS protocol was developed in the early 1980s...
correct: The NFS protocol was developed in the early 1980s...

The other parts of that paragraph use "NAS" in a more generic way, that may include SMB/CIFS. The paragraph could be improved by using NFS throughout and then mentioning the "not-invented-here-syndrome" SMB/CIFS protocol, used e.g. with HyperV. Existence of more obscure protocols (AFS) or newer developments (Ceph) might be hinted at. Anyway, there is not one "NAS protocol".

Anonymous  Jan 20, 2015 
PDF Page 7
p3, p4

Sometimes you italicize 'bash', but sometimes it's plaintext. I suggest monospace (like 'sudo', 'ps', etc.).

Chris Pepper  Feb 27, 2013 
PDF Page 9
p1

Zero Touch Provisioning (ZTP) allows your Arista switch to not only load its configuration from the network, but also from its operating system.
Zero Touch Provisioning (ZTP) allows your Arista switch to not only load its configu?ration from the network, but also its operating system.

Chris Pepper  Feb 27, 2013 
Printed Page 13
last paragraph

Last paragraph describes a situation where e3 sends a message to e1. On the next page the example continues, but it is talking about e3 sending to e2 (not e1).

Sentence on page 13 should read '...when e3 sends a message to e2...'

Daniel Bethke  Jan 16, 2013 
Printed Page 15
figure 2-4

In the example given in the text, the relevant buffers would be output buffers. The figure could be improved by moving the buffers to the output side.

Anonymous  Jan 20, 2015 
Printed Page 21
Top

Hello,

We can read "For fun, the same buffer would hold only -.019 seconds of traffic at 10 megabits."
While it looks like to be "For fun, the same buffer would hold only 0.019 seconds of traffic at 10 gigabits."

Regards,

Jean-Edouard Babin  Nov 16, 2012 
21
1st

I set up my lab with the following IP addresses:
Arista switch: 192.168.1.188
sFlow collector: 192.168.1.100
The first step in configuring sFlow on the switch is to tell the switch where the collector is located. This is done with the sflow destination ip-address port# global configuration command. The port number is optional, and defaults to 6343. Since I?m fine with that port, I?ll just include the IP address:
SW1(config)#sflow destination 192.168.1.188

I would assume the sFlow destination would be the collector. So 192.168.1.100.

saar  May 08, 2013 
PDF Page 21
1st paragraph

Fifty ms is 1/20th of a second, or 0.20 seconds.
Fifty ms is 1/20th of a second, or 0.05 seconds.

Nathan Wolfe  Jan 28, 2014 
PDF Page 23
p3, p4

used to described chips,
used to describe chips,

Chris Pepper  Feb 27, 2013 
Printed Page 24
3rd bullet list

comma used where there should be none:

bad: Easy to implement, newer ASICs
good: Easy to implement newer ASICs

Anonymous  Jan 20, 2015 
Printed Page 33
middle

Hi,

"sports 32 10 Gbps ports" should be "supports 32 10 Gbps ports"

Regards,

Jean-Edouard Babin  Nov 19, 2012 
Printed Page 34
figure 4-5 caption

missing 'r' in 'ports'

wrong: ...dedicated pots highlighted
right: ...dedicated ports highlighted

Anonymous  Jan 20, 2015 
Printed Page 38
last paragraph

The Arista 7500 chassis 48-port 10 Gbps non-blocking modules has one ASIC per...

You should write either "... modules have ..." or "... module has ...", but the combination of "modules" (plural) and "has" (singular) is incorrect.

Anonymous  Jan 20, 2015 
Printed Page 40
"Airflow" section

The description of the meaning of the red and blue handles is not quite correct. it was correct at the time of publishing simply because at that point all of our products had their fans and PSUs on the back of the unit

The precise meaning of the red and blue handles is that blue handles are on the intake side of the unit and red handles are on the exhaust side. This means that on a 7300-series chassis configured for front-to-back airflow, the power supplies (which are on the front) will have blue handles while the fan modules (which are on the back) will have red.

Robert F Merrill  Mar 09, 2014 
PDF Page 42
p3, p4

Both the QSFP+ and the SFP+ support twinax cables
Both QSFP+ and SFP+ support twinax cables

Chris Pepper  Feb 27, 2013 
PDF Page 45
p2

a non-blocking 480 Gbps backplane, which ensures a non- blocking backplane

Chris Pepper  Feb 27, 2013 
PDF Page 79-81
All

The first three pages of the LLDP chapter are incorrect in regards to CDP. The book says that the Arista switch will listen to and understand the CDP advertisements and then put them in its LLDP table; however, this is not the case.

In an email from Gary Donahue:

"An Arista switch will not listen to CDP and add it to the LLDP table. In order to make the Arista switch "hear" the Cisco switch, you need to enable LLDP on the Cisco switch. The Arista switch will listen to CDP when using VMTracer, but it's sort of hidden and not obvious that it's doing it. It's also contained within VMTRacer - there is no CDP table that can be seen."

Kyle Birkeland  Apr 03, 2013 
PDF Page 98
End of Second Paragraph

Arista#boot system flash:/EOS-4.8.1.swi

should be
Arista(config)#boot system flash:/EOS-4.8.1.swi

Hiten Thakkar  Oct 07, 2014 
Printed Page 102
last line of modified code

the last line of new code reads:

ptint ""

but it should read:

print ""

Anonymous  Jan 20, 2015 
PDF Page 104
code example at bottom of page

# Version is same as original - proced with copy
# Version is same as original - proceed with copy

Omar Ali  Apr 15, 2013 
PDF Page 108
p1

the available bandwidth too and from switch C
the available bandwidth to and from switch C

Chris Pepper  Feb 27, 2013 
Printed Page 108
2nd paragraph

"dot1Q" is used to refer to link aggregation), but 802.1Q standardizes VLAN tagging. Link aggregation is standardized in 802.1AX (formerly in 802.3ad, thus link aggregation and LACP are usually known as 802.3ad).

Erik Auerswald  Jul 10, 2015 
Printed Page 110
3rd paragraph

incorrect: "Here's what would happen on a Cisco switch:"
correct: "Here's what would happen on some Cisco switches:"

Only some Cisco switches actually support ISL. A very common example for no ISL support is the Catalyst 2960 series. On that switch series, the mentioned error does not occur.

Erik Auerswald  Jul 10, 2015 
Printed Page 113
1st paragraph

"dot1Q" is used to refer to link aggregation), but 802.1Q standardizes VLAN tagging. Link aggregation is standardized in 802.1AX (formerly in 802.3ad, thus link aggregation and LACP are usually known as 802.3ad).

Erik Auerswald  Jul 10, 2015 
Printed Page 123
numbered list

The first list item has a line break in the word "detail", printing it as two words over two lines ("de\ntail").

Erik Auerswald  Jul 10, 2015 
Printed Page 127
2nd paragraph

not good: "[...] that Cisco 3750s only support STP up to 256 VLANs."
good" "[...] that Cisco 3750s only support up to 256 STP instances."

The limit is for STP instances, which equates VLANs iff some kind of per-VLAN-STP is used. With MST, Cisco supports a higher number of STP instances.

Of course, the number of supported STP instances is model specific.

Erik Auerswald  Jul 10, 2015 
Printed Page 133
1st text paragraph

The description of MST behaviours very clearly shows the author's Cisco background. I'd suggest rewriting the whole paragraph.

MST BPDUs need to be sent out every switch port, without a VLAN tag. The MST BPDU comprises the CST and any configured instance. No MST BPDU is sent with a VLAN tag per the standard. MST BPDUs "must" be sent on any port, including "access ports" in VLANs different from 1 and trunks without any untagged VLAN ("tag native" in Cisco IOS).

MST works very differently from the per-VLAN spanning trees (Cisco's PVSTP family, Alcatel's 1x1 mode, Extreme's EMISTP, and all the more-or-less Cisco PVSTP(+) compatible 3rd party implementations).

I have checked neither Arista's nor the Linux kernel's MST implementations yet, but the standard defines how it should work, and Enterasys Network switches (especially N-Series and S-Series) show how to implement it and mitigate other vendors' mistakes to save the network, too.

Erik Auerswald  Jul 10, 2015 
Printed Page 136
2nd to last paragraph

Only the plus (+) variants of (R)PVST use multicast MAC destination addresses. This often overlooked change, together with untagged CST BPDUs, created some compatibility between Cisco's proprietary default STP and other vendors' standards based default STPs. Thus the + should be added here explicitly.

Erik Auerswald  Jul 10, 2015 
Printed Page 136
last paragraph

wrong: "[...] Cisco enhanced the MST protocol [...]"
right: "[...] Cisco uses a non-standard-compliant MST implementation [...]"

Erik Auerswald  Jul 10, 2015 
Printed Page 137
3rd text paragraph

wrong: "With MST, BPDUs are only ever sent over VLAN1 (by default), [...]"
right: "With MST, BPDUs are only ever sent untagged, [...]"

Erik Auerswald  Jul 10, 2015 
Printed Page 155
2nd paragraph

VRRP is called an "open source protocol", while it is just an "open" standard. In fact, VRRP is patent-encumbered, which lead to the development of the open source, even free software, CARP protocol and implementations.

Please remove the word "source" from between "open" and "protocol". Thanks.

Erik Auerswald  Jul 10, 2015 
Printed Page 169
figure 14-4

It would be better to attach the servers directly to SW1 resp. SW2, as in an STP enabled layer 2 network (which would connect all the servers in VLAN 99) usually the remote end of the link from Sw2 et10 would be blocked and thus not used at all.

VARP is very useful for an MLAG configuration, e.g. the server 20.20.20.20 would use both SW1 and SW2 as default gateway, depending on the hash for each individual flow.

The picture is not "wrong", it is just a bit too vague and probably leads people to wrong ideas (just as most pictures showing setups with Cisco's GLBP).

Erik Auerswald  Jul 10, 2015 
Printed Page 170
tech tip at top of page

The ARP responses for the virtual IP from both of the switches are identical. The layer 2 network determines the path used by a frame from a server to reach one of the VARP enabled gateways. The first gateway reached by the frame will process it for layer 3 forwarding. Thus the peer link is never used to reach the "active" gateway with VARP, as both peers are active. This is an important difference compared to standard VRRP.

[Of course other vendors implement non-standard VRRP extensions to the same effect, e.g. Enterasys Networks' "fabric routing". For a good laugh look at Extreme Networks' active-active VRRP configuration -- they create something similar to VARP by creating a split-brain VRRP situation.]

Erik Auerswald  Jul 10, 2015 
Printed Page 170
2nd text paragraph

Please delete "in routing, nor can it be used".

The VIP can be used for static routing.

Erik Auerswald  Jul 10, 2015 
Printed Page 170
last line of page

The "rule for making up MAC addresses" is to set the second lowest bit in the first octet to 1. So you could change the Arista OUI used to "02:1c:73" to create a locally administered MAC address (02:1c:73:00:00:99). It would probably work to just use 02:00:00:00:00:99.

Erik Auerswald  Jul 10, 2015 
Printed Page 173
1st text paragraph

wrong: "[...] show ip virtual router [...]"
right: "[...] show ip virtual-router [...]"

The hyphen is missing from "virtual-router".

Erik Auerswald  Jul 10, 2015 
PDF Page 175
p2

the way you?d expect them too.
the way you?d expect them to.

Chris Pepper  Feb 27, 2013 
Printed Page 186
1st text paragrph (after the list starting the page)

The text omits the crucial part of _unsupported_.

wrong: The same document also lists the following:"
right: "The same document also lists the following as not supported:"

Currently both egress ACLs, and IPv6 and MAC ACLs are supported by Arista (depending on hardware support), but the text references a specific EOS version.

Erik Auerswald  Jul 10, 2015 
Printed Page 194
2nd text paragraph

VRRP is an IP based protocol with protocol number 112, but the text claims it is _not_ part of IP and would use protocol number 118. The author might have been confused by the IP multicast address used, which is 224.0.0.18.

Erik Auerswald  Jul 10, 2015 
Printed Page 199,200
bottom of 199 start of 200

End of page 199 and start of 200.

Where you say there are two fields we care about in his chapter. You say we care about CFI and VID. But you defined CFI as relating to token ring. I think you meant pcp instead. Either you meant pcp or the definitions on page 199 are wrong.

Craig Gleason  Nov 28, 2012 
Printed Page 199
definition list at bottom of page, 1st item

The definition of "TPID" has a spurious "0v" in front of "0x8100", which should be deleted.

Erik Auerswald  Jul 10, 2015 
PDF Page 211
p1

This can be altered with the 􏰣􏰅􏰑􏰓􏰕􏰂􏰓􏰄􏰥􏰍􏰙􏰚􏰁􏰘􏰚􏰑􏰄 􏰙􏰚􏰁􏰘􏰚􏰑􏰄 􏰅􏰐􏰚 tx-queue command.
This can be altered with the bandwidth percent percentage tx-queue command.
On the page, "percent age" is wrapped without a hyphen.

Chris Pepper  Feb 27, 2013 
Printed, Page 258
1st Paragraph under "Configuring sFlow"

The correct URL for sFlowTrend is http://www.inmon.com

Harish Dommathamari  Jun 08, 2016 
PDF Page 263
2

VM Tracer can be given a range of VLANs with which it can add,
VM Tracer can be given a range of VLANs which it can add,

Chris Pepper  Feb 27, 2013 
PDF Page 266
1

The trunk between the switched (port-channel 20 on both sides)
The trunk between the switches (port-channel 20 on both sides)

Chris Pepper  Feb 27, 2013 
PDF Page 271
3

Luckily, we can limit the number of VLANs for which VM Tracer is allowed to change.
Luckily, we can limit the VLANs which VM Tracer is allowed to change.
It's not just *quantity* -- it's selection of particular VLANs.

Chris Pepper  Feb 27, 2013 
PDF Page 276
1

but it get?s worse.
but it gets worse.

Chris Pepper  Feb 27, 2013 
Printed Page 284
last line on page

Missing space in Linux command:

wrong: "ps-ef r"
right: "ps -ef r"

Erik Auerswald  Jul 10, 2015 
PDF Page 287
Intro

It would be worth mentioning in the introduction that 'tcpdump' and 'wireshark' use a standard file format and files are intercompatible, so you can capture data with 'tcpdump' on the switch and analyze it at leisure on a graphical workstation.

Chris Pepper  Feb 27, 2013 
Printed Page 291
1st text paragraph

The text in this first paragraph uses "STP.capture" as file name, but all the examples below use "GAD.capture". The text should be changed to use "GAD.capture" as well.

Erik Auerswald  Jul 10, 2015 
Printed Page 295
1st paragraph

Mispelling of MLAG where it says MALG interface (VLAN 4094)

Craig Gleason  Nov 29, 2012 
Printed Page 306-307
code block on 306 and dhcp config on 307

MAC address shown in code block on 306 does not match MAC address shown in dhcp options section on 307

MAC on 306 is 001c.7308.80.ac (00:1c:73:08:80:ac in dhcp format)

MAC on 307 is 001c.7308.80ae (00:1c:73:08:80:ae in dhcp format)

They should match.

Craig Gleason  Nov 29, 2012 
Printed Page 312
tech tip at top of page, 1st paragraph therein

there is a spurious closing parenthesis after VMTracer that should be deleted:

wrong: "[...] a VM discovered with VMTracer) is moved."
right: "[...] a VM discovered with VMTracer is moved."

Erik Auerswald  Jul 10, 2015 
PDF Page 317
3

did not effect any
did not affect any

Chris Pepper  Feb 27, 2013 
Printed Page 359
2nd paragraph

In 2nd paragraph, the commaned mentioned is the reboot now command. I believe the correct command (and one shown in figure 29-16) is actually reload now.

Craig Gleason  Nov 30, 2012 
Printed Page 367
1st paragraph

The load values are NOT given in percent, they are given in a fractional number representing the number of concurrently runnable processes. As such the description given is just wrong, the percentages show are totally misleading.

It is correct that this number does not directly tell you if a switch is "overloaded" (no pun intended ;) or not, but it does help to know what it means.

Erik Auerswald  Jul 10, 2015 
PDF Page 380
3

other vendor?s switches.
other vendors? switches.

Chris Pepper  Feb 27, 2013 
PDF Page 381
4

hardcoded 1,000 Gbps/full duplex
hardcoded 1,000 Mbps/full duplex

Chris Pepper  Feb 27, 2013 
Printed Page 383
Trunk groups section, paragraph 3

There is a mix of the trunk group name used through the Trunk groups example. Some times the trunk group name 'Servers' is referenced and other times the trunk group 'Leelu' is referenced. I believe all references to 'Leelu' should be replaced with 'Servers' on pages 383-386

Tom Leach  Sep 12, 2013 
Printed Page 383
5th normal text paragraph

wrong: "We've already see it in use for the MLAG peer, [...]"
right: "We've already seen it in use for the MLAG peer, [...]"

Missing n, that is "see" used where "seen" needs to be.

Erik Auerswald  Jul 10, 2015 
Printed Page 383
2nd to last text paragraph

wrong: "[...] apply the trunk group Servers [...]"
right: "[...] apply the trunk group Leelu [...]"

This one sentence uses the trunk group "Servers", while all the examples and the text on the same page use "Leelu" as trunk group name. It might be correct to implement the change suggested by Tom Leach and replace Leelu with Server, but I like the movie reference and would prefer to change "Servers" to "Leelu" (or maybe "Leeloo" for the pedantic ;).

Erik Auerswald  Jul 10, 2015