Arista Eos Rn-4.19.1f

Arista Networks EOS 4.19.1F Release Notes Version: 1.0

Arista Networks Confidential

Table of Contents EOS 4.19.1F Release Highlights New Platforms and Hardware SNMP MIBs SNMP Traps Supported Hardware Picking compatible versions for MLAG ISSU Upgrade/Downgrade Release 4.19.X MLAG ISSU Table Release 4.18.X MLAG ISSU Table Release 4.17.X MLAG ISSU Table Release 4.16.X MLAG ISSU Table Release 4.15.X MLAG ISSU Table Enhancements in 4.19.1F Resolved Caveats in 4.19.1F General Known Software Caveats in 4.19.1F Layer3 Known Software Caveats in 4.19.1F DCS-7160 Known Software Caveats in 4.19.1F DCS-7500 Known Software Caveats in 4.19.1F Multicast Known Software Caveats in 4.19.1F DCS-7300 Known Software Caveats in 4.19.1F General Limitations and Restrictions in 4.19.1F DCS-7300 Limitations and Restrictions in 4.19.1F Conditions and Impacts

Arista Networks Confidential 2

EOS 4.19.1F Release Highlights New Platforms and Hardware • • • •

DCS-7500R2M-36CQ-LC DCS-7280CR2A-60-F DCS-7020TRA-48-F DCS-7020TRA-48-R

SNMP MIBs • SNMPv2, SNMPv3 • RFC 3635 EtherLike-MIB ◦ obsoletes RFCs 1650, 2358, 2665 • RFC 3418 SNMPv2-MIB ◦ obsoletes RFCs 1450, 1907 • RFC 2863 IF-MIB ◦ obsoletes RFCs 1229, 1573, 2233 • RFC 2864 IF-INVERTED-STACK-MIB • RFC 2096 IP-FORWARD-MIB ◦ obsoletes RFC 1354 • ARISTA-SW-IP-FORWARD-MIB ◦ IPv4 only • RFC 4363 Q-BRIDGE-MIB • RFC 4188 BRIDGE-MIB • ARISTA-BRIDGE-EXT-MIB • RFC 2013 UDP-MIB ◦ obsoletes RFC 1213 • RFC 2012 TCP-MIB ◦ obsoletes RFC 1213 • RFC 2011 IP-MIB ◦ obsoletes RFC 1213 • HOST-RESOURCES-MIB • LLDP-MIB • LLDP-EXT-DOT1-MIB • LLDP-EXT-DOT3-MIB • ENTITY-MIB • ENTITY-SENSOR-MIB • ENTITY-STATE-MIB • RMON-MIB ◦ rmonEtherStatsGroup • RMON2-MIB ◦ rmon1EthernetEnhancementGroup • HC-RMON-MIB ◦ etherStatsHighCapacityGroup Arista Networks Confidential 3

• RFC 3636 MAU-MIB ◦ ifMauDefaultType and ifMauAutoNegStatus are writeable • SNMP-TLS-TM-MIB ◦ RFC 6353 • SNMP-TSM-MIB ◦ RFC 5591 • ARISTA-ACL-MIB • ARISTA-SNMP-TRANSPORTS-MIB ◦ 7050X2, 7060X, 7260X,7320X, 7060X2, 7160,7020R • ARISTA-SMI-MIB ◦ 7050X2, 7060X, 7260X,7320X, 7060X2, 7160,7020R • ARISTA-PRODUCTS-MIB ◦ 7050X2, 7060X, 7260X,7320X, 7060X2, 7160,7020R • ARISTA-QUEUE-MIB ◦ Excluding 7150 • RFC 4273 BGP4-MIB • RFC 4750 OSPF-MIB • ARISTA-CONFIG-COPY-MIB • ARISTA-CONFIG-MAN-MIB • ARISTA-REDUNDANCY-MIB ◦ 7500R, 7500E, 7300X, 7050X2, 7060X, 7260X,7320X, 7060X2, 7160,7020R • MSDP-MIB • PIM-MIB • IGMP-MIB • IPMROUTE-STD-MIB • VRRPV2-MIB • ARISTA-QOS-MIB • ARISTA-ENTITY-SENSOR-MIB • ARISTA-BGPV4V2-MIB • ARISTA-VRF-MIB • ARISTA-DAEMON-MIB • ARISTA-ECN-COUNTER-MIB ◦ 7500R, 7500E, 7280R, 7280E • ARISTA-IF-MIB ◦ 7050X2, 7060X, 7260X,7320X, 7060X2, 7160,7020R • ARISTA-MAU-MIB ◦ 7050X2, 7060X, 7260X,7320X, 7060X2, 7160,7020R • ARISTA-PFC-MIB ◦ 7050X2, 7060X, 7260X,7320X, 7060X2, 7160,7020R

SNMP Traps • RFC 2863 IF-MIB ◦ linkUp, linkDown • LLDP-MIB ◦ lldpRemTablesChange Arista Networks Confidential 4

• RFC 3418 SNMPv2-MIB ◦ coldStart • NET-SNMP-AGENT-MIB ◦ nsNotifyRestart • ENTITY-MIB ◦ entConfigChange • ENTITY-STATE-MIB ◦ entStateOperEnabled, entStateOperDisabled • OSPF-MIB ◦ ospfNbrStateChange, ospfIfConfigError, ospfIfAuthFailure, ospfIfStateChange • BGP4-MIB ◦ bgpEstablished, bgpBackwardTransition • ARISTA-REDUNDANCY-MIB ◦ aristaRedundancySwitchOverNotif only for: 7500R, 7500E, 7300X, 7050X2, 7060X, 7260X,7320X, 7060X2, 7160,7020R • ARISTA-CONFIG-MAN-MIB ◦ aristaConfigManEvent • SNMPv2-MIB ◦ authenticationFailure • VRRPv2-MIB ◦ vrrpTrapNewMaster, vrrpTrapAuthFailure

Supported Hardware Fixed Systems And Accessories • DCS-7010T-48 • DCS-7010T-48-DC

• DCS-7150S-24-CL-R • DCS-7150S-24-CL-SSD-F

• DCS-7280TR-48C6-M-R • DCS-7280CR-48-F

• DCS-7050SX-64-F • DCS-7050SX-64-R

• DCS-7150S-24-CL-SSD-R • DCS-7150S-52-CL-F

• DCS-7280CR-48-D • DCS-7280CR-48-DC-F

• DCS-7050SX-64-SSD-F • DCS-7050SX-64-SSD-R

• DCS-7150S-52-CL-R • DCS-7150S-52-CL-SSD-F

• DCS-7280SR2-48YC6-F • DCS-7280SR2-48YC6-R

• DCS-7050SX-72-F • DCS-7050SX-72-R

• DCS-7150S-52-CL-SSD-R • DCS-7150S-64-CL-F

• DCS-7280SR2-48YC6-M-F • DCS-7280SR2-48YC6-M-R


• DCS-7150S-64-CL-R • DCS-7150-64-CL-SSD-F

• DCS-7280SRA-48C6-F • DCS-7280SRA-48C6-R

• DCS-7050SX-96-F • DCS-7050SX-96-R

• DCS-7150-64-CL-SSD-R • DCS-7160-32CQ-F

• DCS-7280TRA-48C6-F • DCS-7280TRA-48C6-R


• DCS-7160-32CQ-R • DCS-7160-32CQ-SSD-F

• DCS-7280SRA-48C6-M-F • DCS-7280SRA-48C6-M-R

• DCS-7050SX-128-F • DCS-7050SX-128-R

• DCS-7160-32CQ-SSD-R • DCS-7160-48YC6-F

• DCS-7280SR2A-48YC6-F • DCS-7280SR2A-48YC6-R


• DCS-7160-48YC6-R • DCS-7160-48TC6-F

• DCS-7280SR2A-48YC6-M-F • DCS-7280SR2A-48YC6-M-R

• DCS-7050SX2-72Q-F • DCS-7050SX2-72Q-R

• DCS-7160-48TC6-R • DCS-7250QX-64-F

• DCS-7020TR-48-F • DCS-7020TR-48-R


• DCS-7050SX2-128-F • DCS-7050SX2-128-R

• DCS-7250QX-64-R • DCS-7250QX-64-SSD-F

• PWR-1100AC-F • PWR-1100AC-R

• DCS-7050TX-48-F • DCS-7050TX-48-R

• DCS-7250QX-64-SSD-R • DCS-7260CX-64-F

• PWR-1900AC-F • PWR-1900-DC-F

• DCS-7050TX-48-SSD-F • DCS-7050TX-48-SSD-R

• DCS-7260CX-64-R • DCS-7260CX-64-SSD-F

• PWR-1900-DC-R • PWR-745AC-F

• DCS-7050TX-64-F • DCS-7050TX-64-R

• DCS-7260CX-64-SSD-R • DCS-7260QX-64-F

• PWR-745AC-R • PWR-750AC-F

• DCS-7050TX-64-SSD-R • DCS-7050TX-64-SSD-F

• DCS-7260QX-64-R • DCS-7260QX-64-SSD-F

• PWR-750AC-R • PWR-650AC

• DCS-7050TX-72-F • DCS-7050TX-72-R

• DCS-7260QX-64-SSD-R • DCS-7280SE-64-F



• DCS-7280SE-64-R • DCS-7280SE-68-F

• PWR-460DC-F • PWR-460DC-R

• DCS-7050TX-96-F • DCS-7050TX-96-R

• DCS-7280SE-68-R • DCS-7280SE-72-F



• DCS-7280SE-72-R • DCS-7280QR-C36-F

• PWR-500-DC-F • PWR-500-DC-R

• DCS-7050TX-128-F • DCS-7050TX-128-R

• DCS-7280QR-C36-R • DCS-7280QR-C36-M-F

• FAN-7002-F • FAN-7002-R

• DCS-7050TX-128-SSD-R • DCS-7050TX-128-SSD-F

• DCS-7280QR-C36-M-R • DCS-7280QR-C72-F

• FAN-7002H-F • FAN-7000-F

• DCS-7050TX2-128-F • DCS-7050TX2-128-R

• DCS-7280QR-C72-R • DCS-7280QR-C72-M-F

• FAN-7000-R • FAN-7000H-F

• DCS-7060CX-32S-F • DCS-7060CX-32S-R

• DCS-7280QR-C72-M-R • DCS-7280SR-48C6-F

• FAN-7000H-R • DCS-7280TRA-48C6-M-F

• DCS-7060CX-32S-SSD-F • DCS-7060CX-32S-SSD-R

• DCS-7280SR-48C6-R • DCS-7280SR-48C6-M-F

• DCS-7280TRA-48C6-M-R • DCS-7280QRA-C36S-F

• DCS-7060CX2-32S-F • DCS-7060CX2-32S-R

• DCS-7280SR-48C6-M-R • DCS-7280TR-48C6-F

• DCS-7280QRA-C36S-R • DCS-7280CR2A-60-F

• DCS-7150S-24-F • DCS-7150S-24-R

• DCS-7280TR-48C6-R • DCS-7280TR-48C6-M-F

• DCS-7020TRA-48-F • DCS-7020TRA-48-R

• DCS-7150S-24-CL-F

Modular Systems and Accessories • DCS-7508 • DCS-7504

• 7500RM-36CQ-LC • 7500R-48S2CQ-LC

• 7316X-FM • 7320X-32C-LC

• DCS-7504N • DCS-7508N

• 7500R2-36CQ-LC • 7500R2-18CQ-LC

• 7324X-FM-F • 7328X-FM-F

• DCS-7512N • 7500E-SUP

• 7504E-FM • 7508E-FM

• 7304-S-FAN • 7308-S-FAN

• 7500E-SUP-D • 7500-SUP2

• 7508R-FM • 7504R-FM

• 7500R2AK-48YCQ-LC • PWR-2900AC

• 7500-SUP2-D • 7500E-36Q-LC

• 7512R-FM • DCS-7304

• PWR-3KT-AC-RED • PWR-3K-DC-RED

• 7500E-72S-LC

• DCS-7308

• PWR-3K-AC-F


• 7500E-48S-LC • 7500E-48T-LC

• DCS-7316 • 7300-SUP

• PWR-3K-AC-R • PWR-2700-DC-F

• 7500E-12CM-LC • 7500E-6C2-LC

• 7300-SUP-D • 7300X-32Q-LC

• PWR-2700-DC-R • FAN-7002-F

• 7500E-12CQ-LC • 7500R-8CFPX-LC

• 7300X-64S-LC • 7300X-64T-LC

• FAN-7002-R • 7500R2A-36CQ-LC

• 7500R-36CQ-LC • 7500R-36Q-LC

• 7304X-FM • 7308X-FM

• 7500R2AK-36CQ-LC • 7500R2M-36CQ-LC

• 100GBASE-SR4 • 100GBASE-AOC

• 40GBASE-PLR4 • 40GBASE-PLRL4

• 10GBASE-DWDM • 10GBASE-DWDM-ZR

• 100GBASE-CR4 • 100GBASE-LR4

• 40GBASE-SR4 • 40GBASE-ER4

• 10GBASE-DWDM-T • 10GBASE-ZR

• 100GBASE-LRL4 • 100GBASE-CWDM4

• 40GBASE-XSR4 • 40GBASE-LRL4

• 10GBASE-ER • 10GBASE-LR

• 100GE-DWDM2 • CFP2-100GBASE-LR4

• 40GBASE-UNIV • 40GBASE-SRBD

• 10GBASE-LRL • 10GBASE-SR

• CFP2-100GBASE-ER4 • CFP2-100GBASE-XSR10

• 25GBASE-CR • 25GBASE-SR

• 10GBASE-SRL • 1000BASE-SX

• CFPX-100G-DWDM • CFPX-200G-DWDM

• 25GBASE-AOC • 25GBASE-LR

• 1000BASE-LX • 1000BASE-BX10-D

• 40GBASE-AOC • 40GBASE-CR4

• 25GBASE-LRL • 10GBASE-AOC

• 1000BASE-BX10-U • 1000BASE-T

• 40GBASE-LR4

• 10GBASE-CR

Transceivers


Picking compatible versions for MLAG ISSU Upgrade/ Downgrade Upgrading/downgrading using the MLAG ISSU procedure, from release EOS-A.B.C to release EOS-D.E.F can be done in one or multiple phases. 1. If A.B.C is a compatible EOS version for D.E.F (refer to the A.B.X or D.E.X release table) then you can directly do MLAG ISSU using the D.E.F version by following the MLAG ISSU procedure. (A.B.C -> D.E.F). Example: The customer wants to upgrade from 4.12.10 to 4.15.4F. In the Release 4.15.X MLAG ISSU Table, the 4.15.4F version has 4.12.10 as a MLAG ISSU compatible version, so customer can directly do an MLAG ISSU upgrade from 4.12.10 to 4.15.4F. 2. If A.B.C doesn't exist in the list of compatible versions for D.E.F (refer to the A.B.X or D.E.X release table), then look for the A.B.Y version in D.E.F's ISSU compatible EOS versions, do MLAG ISSU upgrade/downgrade from A.B.C to A.B.Y and then do MLAG ISSU upgrade/downgrade to D.E.F by following MLAG ISSU procedure (A.B.C -> A.B.Y -> D.E.F). Here, Y can be greater than or less than C. Example: The customer wants to upgrade from 4.14.1F to 4.15.4F. In Release 4.15.X MLAG ISSU Table 4.15.4F version has 4.14.9F as MLAG ISSU compatible version and Release 4.14.X MLAG ISSU Table 4.14.9F has 4.14.1F as MLAG ISSU compatible version, so customer can do MLAG ISSU upgrade from 4.14.1F to 4.14.9F and then to 4.15.4F. 3. Look for an D.E.Z version that's MLAG ISSU compatible with both A.B.C and D.E.F. Do MLAG ISSU upgrade/downgrade from A.B.C to D.E.Z and then do MLAG ISSU upgrade/downgrade to D.E.F (A.B.C -> D.E.Z -> D.E.F). Here Z can be greater than or less than F. Example: The customer wants to upgrade from 4.14.7F to 4.15.4F. In Release 4.15.X MLAG ISSU Table 4.15.0F version has 4.14.7F as MLAG ISSU compatible version and 4.15.4F has 4.15.0F as MLAG ISSU compatible version, so customer can do MLAG ISSU upgrade from 4.14.7F to 4.15.0F and then to 4.15.4F. 4. Look for an G.H.I version that's MLAG ISSU compatible with both A.B.C and D.E.F. Do an MLAG ISSU upgrade/downgrade from A.B.C to G.H.I and then do Mlag ISSU upgrade/downgrade to D.E.F (A.B.C -> G.H.I -> D.E.F). 5. In a highly unlikely scenario, if any of the above procedures are not acceptable, MLAG ISSU can be achieved by picking multiple intermediate versions.


Release 4.19.X MLAG ISSU Table EOS Version

Mlag ISSU compatible EOS versions

4.19.0F

4.16.6M-4.16.10M, 4.16.7FX-MLAGISSU-TWO-STEP, 4.16.8FX-MLAGISSUTWO-STEP, 4.17.0F-4.17.3F, 4.17.4M, 4.18.0F, 4.18.1F-4.18.4F, 4.18.1.1F

4.19.1F

4.16.6M-4.16.12M, 4.16.7FX-MLAGISSU-TWO-STEP, 4.16.8FX-MLAGISSUTWO-STEP, 4.17.0F-4.17.3F, 4.17.4M-4.17.7M, 4.18.0F-4.18.4F



4.18.0F

4.16.6M-4.16.9M, 4.16.7FX-MLAGISSU-TWO-STEP, 4.16.8FX-MLAGISSUTWO-STEP, 4.17.0F-4.17.3F

4.18.1F

4.16.6M-4.16.10M, 4.16.7FX-MLAGISSU-TWO-STEP, 4.16.8FX-MLAGISSUTWO-STEP, 4.17.0F-4.17.3F, 4.17.4M, 4.18.0F

4.18.1.1F

4.16.6M-4.16.10M, 4.16.7FX-MLAGISSU-TWO-STEP, 4.16.8FX-MLAGISSUTWO-STEP, 4.17.0F-4.17.3F, 4.17.4M, 4.18.0F, 4.18.1F

4.18.2F

4.16.6M-4.16.10M, 4.16.7FX-MLAGISSU-TWO-STEP, 4.16.8FX-MLAGISSUTWO-STEP, 4.17.0F-4.17.3F, 4.17.4M, 4.18.0F, 4.18.1F, 4.18.1.1F

4.18.3F

4.16.6M-4.16.10M, 4.16.7FX-MLAGISSU-TWO-STEP, 4.16.8FX-MLAGISSUTWO-STEP, 4.17.0F-4.17.3F, 4.17.4M, 4.18.0F, 4.18.1F, 4.18.1.1F 4.18.2F

4.18.4F

4.16.6M-4.16.10M, 4.16.7FX-MLAGISSU-TWO-STEP, 4.16.8FX-MLAGISSUTWO-STEP, 4.17.0F-4.17.3F, 4.17.4M, 4.18.0F, 4.18.1F, 4.18.1.1F, 4.18.2F, 4.18.3F



4.17.0F

4.16.6M

4.17.1F

4.16.6M-4.16.7M, 4.16.7FX-MLAGISSU-TWO-STEP, 4.17.0F

4.17.2F


4.17.3F


4.17.4M





4.16.6M

4.16.7FX-MLAGISSU-TWO-STEP

4.16.7M

4.16.6M, 4.16.7FX-MLAGISSU-TWO-STEP

4.16.8M

4.16.6M-4.16.7M, 4.16.7FX-MLAGISSU-TWO-STEP - 4.16.8FX-MLAGISSUTWO-STEP

4.16.9M


4.16.10M


4.16.7FXMLAGISSU-TWOSTEP

4.9.11, 4.10.8.1, 4.11.12, 4.12.11, 4.13.16M, 4.14.14M, 4.15.0F-4.15.4F, 4.15.5M-4.15.7M, 4.16.6M, 4.16.7M, 4.17.0F

4.16.8FX4.9.11, 4.10.8.1, 4.11.12, 4.12.11, 4.13.16M, 4.14.15M, MLAGISSU-TWO4.15.0F-4.15.4F, 4.15.5M-4.15.8M, 4.16.6M-4.16.8M, 4.17.0F-4.17.1F STEP



4.15.0F

4.9.11, 4.10.8.1, 4.11.11, 4.12.10, 4.13.12M, 4.14.7M

4.15.1F

4.9.11, 4.10.8.1, 4.11.11, 4.12.10, 4.13.12M, 4.14.7M, 4.15.0F

4.15.2F

4.9.11, 4.10.8.1, 4.11.11, 4.12.10, 4.13.12M, 4.14.9M, 4.15.0F-4.15.1F

4.15.3F

4.9.11, 4.10.8.1, 4.11.11, 4.12.10, 4.13.12M, 4.14.9M, 4.15.0F-4.15.2F

4.15.4F

4.9.11, 4.10.8.1, 4.11.11, 4.12.10, 4.13.12M, 4.14.9M, 4.15.0F-4.15.3F

4.15.5M

4.9.11, 4.10.8.1, 4.11.11, 4.12.10, 4.13.12M, 4.14.10M, 4.15.0F-4.15.4F

4.15.6M

4.9.11, 4.10.8.1, 4.11.12, 4.12.11, 4.13.15M, 4.14.12M, 4.15.0F-4.15.4F, 4.15.5M

4.15.7M

4.9.11, 4.10.8.1, 4.11.12, 4.12.11, 4.13.16M, 4.14.13M, 4.15.0F-4.15.4F, 4.15.5M-4.15.6M

4.15.8M

4.9.11, 4.10.8.1, 4.11.12, 4.12.11, 4.13.16M, 4.14.13M, 4.15.0F-4.15.4F, 4.15.5M-4.15.7M

4.15.9M

4.9.11, 4.10.8.1, 4.11.12, 4.12.11, 4.13.16M, 4.14.15M, 4.15.0F-4.15.4F,4.15.5M-4.15.8M

4.15.10M

4.9.11, 4.10.8.1, 4.11.12, 4.12.11, 4.13.16M, 4.14.16M, 4.15.0F-4.15.4F,4.15.5M-4.15.9M Arista Networks Confidential 10

Enhancements Added in 4.19.1F • Added Uni-directional link support on DCS-7500R, DCS-7280R, DCS-7500R2, DCS-7280R2 platforms. (122069) • Rate limiting for mirroring sessions is now supported on selected platforms. (168523) • Added support for the DCS-7500R2M-36CQ-LC linecard (188508) • On the DCS-7260CX and DCS-7320X series, VXLAN routing is now supported. (196926) • Enabled hardware fpga error action log across supported fixed and modular and default to log instead of power cycle (201447) • On the DCS-7280E, DCS-7500E, DCS-7280R and DCS-7500R series, IPv6 ACL is now supported in filtering for ingress mirroring. (204837) • On DCS-7050X, DCS-7060CX, DCS-7250QX-64, and DCS-7300X series switches, IPv6 ACL is now supported for filtered ingress mirroring. (205389) • Support for LDP Pseudowires on Jericho+. (205822)

Resolved Caveats in 4.19.1F • OSPF hello packets might not get transmitted (on interfaces without virtual IP) when virtual IP is configured on the system. The workaround is to make interfaces with virtual IPs passive for OSPF to work on the other interfaces. (115041) • On the DCS-7150S, AH and ESP unicast packets are flooded to all the ports of VLAN despite the destination MAC address being learned. (158993) • On DCS7020 series, if a parity error occurs in a table that is used for multicast (ex: TAR_ECC_ParityErrInt[0]), the table would not be corrected. This may cause unexpected behavior or traffic drops for multicast traffic. This can be observed by running "show platform fap interrupts". A workaround would be to restart the SandMcast agent, thus refreshing all of the multicast tables. (160758) • On an MLAG topology consisting of DCS-7050 series switches, when there are large number of dynamic MAC addresses that can result in hash collisions in the hardware MAC address table, the switches may never age out the dynamic MAC addresses and may cause high CPU utilization of Strata agent.


The suggested workaround is to clear the dynamic MAC addresses from the hardware. (164352) • When the system has a very large number of VXLAN MAC entries, if there is simultaneous substantial churn in VTEPs and corresponding remote MAC entries, platform L2 agent can restart unexpectedly causing momentary disruption in overlay traffic. (165954) • If BFD session is administratively 'shutdown' either locally or on the peer, BFD clients may treat it as BFD session 'Down' instead of 'AdminDown', and incorrectly tear down client sessions. This will cause traffic disruption. (166024) • On the DCS-7508 and DCS-7512 series, a kernel crash may occur if an event like a VLAN related configuration change or linecard removal takes place when there are many control plane bound packets. (169255) • On the DCS-7280, DCS-7280R, DCS-7500E, DCS-7500R, during Config Replace, if there are two large ACLs being applied, each of which exhausts the TCAM, the switch remains in error state after Config Replace and in this state no ACL can be removed or applied. (174815) • On the DCS-7050X2 series, on configurations with large egress ACLs, the forwarding slice agent may restart continuously. The workaround is to reduce the egress ACL utilization. (177071) • On DCS-7250QX and DCS-7300 series, when Vxlan is encapsulated, if a packet needs to be forwarded to a remote host ( route + encap or decap + route + encap ), if the VTEP is reachable through an ECMP route in the underlay and any of the underlay nexthops in the ECMP has unlearnt mac, the packets that are hashed to the nexthop will get dropped. (177580) • BGP may not advertise some prefixes on a switch with internet scale routes in the presence of "update wait-install" and "out-delay" configuration (179246) • On the DCS-7300X, DCS-7500E and the DCS-7500R series, repeated power cycle of one or more linecards over an extended period of time can cause Sysdb to restart. (179837) • If a BGP session is established using secondary IP address or a VRRP virtual address, that BGP session will flap when any layer 3 interface in that VRF goes down. (180082) • On the DCS-7010, DCS-7050X, DCs-7060X. DCS-7250X, DCS-7260X and the DCS-7300X series, the StrataLag agent may crash when a LAG member is removed from running configuration. (184185)


• On the DCS-7250X, DCS-7260X, DCS-7050X, DCS-7060X, and DCS-7300X series, if MLAG fast redirection is configured and activated by a MLAG link, the StrataLag agent crashes if it is restarted. (185054) • If multi-agent mode is configured for routing protocols including BGP, under rare conditions, a route may be installed in the Rib with incorrect attributes. The problem is triggered when the best-path attributes (preference, MED metric, internal or external, etc) change for a route, while that route is being tracked for Inplace Adjacency Replace (IAR) processing due to an earlier session or next-hop change event. The route attributes in the Rib reflect the earlier sate of the path, even as the forwarding information is updated. A future update to the route (bestpath or attribute change) after completion of the IAR phase, will also update the attributes of the route in the Rib table. (187263) • Enabling the VMTracer agent and connecting it to a vCenter instance can cause VMTracer to trigger an unexpected system reset. The workaround is to disable the VMTracer agent. (188147) • After a BGP peer session flap, random memory corruption may happen and may lead to Rib agent restart or some undefined behavior (188693) • On the DCS-7280R and DCS-7500R series, if data flows with multiple traffic classes are sent to an egress port, the lower traffic class packets might get dropped on egress chip. (189747) • Rib agent may restart unexpectedly when BGP has received more than 1024 paths for an ECMP prefix and "show ip bgp " command is executed. "show ip bgp [vrf all|default]" command shall also fail, but not result in Rib agent restart. (190191) • On the DCS-7280E, DCS-7280R, DCS-7500E, and DCS-7500R series in Tap Aggregation mode, configuring a monitor session whose source is a tool port with egress time stamping enabled can cause the SandFap agent to restart unexpectedly. The workaround is to avoid using tool ports with egress time stamping enabled as source interfaces in any monitor session. (190197) • When using the multi-agent routing model, a dynamic-peer rapid BGP session teardown and immediate re-establishment may prevent subsequent configuration changes affecting that peer from being applied. (190209) • With the multi-agent routing model, the Bgp agent may restart unexpectedly when generating updates for paths whose serialized length exceeds the BGP update maximum size (4096 octets). AS path prepend, peer specific outbound policy, export extended communities, or four-byte AS configuration may Arista Networks Confidential 13

contribute to the size of the serialized path attributes. This issue is typically caused by a mis-configuration of the network disabling the AS path loop check resulting in looped routes with very long AS paths. (190420) • On DCS-7160 series of switches, when IGMP Snooping and PIM are enabled on a vlan, the DCS-7160 switch must be the only PIM router in that vlan. Workaround would be to disable IGMP snooping (191851) • On the 7160 series switches, if VRRP is configured and the XpMact agent is restarted, the XpMact agent will restart continuously, affecting MAC learning. (192074) • On DCS-7160 series switches, large configurations edited many times can cause the XpMact and XpCentral agents to unexpectedly restart. The only workaround is to reboot the switch. (192082) • On the DCS-7508N, hot swapping a power supply may cause the Fru agent to crash continuously (192144) • Extensions may not be property re-installed at boot time if the 'force' keyword was used during initial installation. To work around this issue the extension can be installed after boot uing the 'force' keyword. (192362) • On the DCS-7050X series, if QoS trust mode is configured as DSCP, the Strata agent may restart when a code point is mapped to a traffic class whose queues are disabled by the PFC watchdog. This may lead to some traffic loss. (192936) • On the DCS-7280QR-C72, ports 11,13,15,17,19,21,23,25,47,49,51,53,55,57,59,61 may fail to link up when using 100G optical transceivers. (193378) • On the DCS-7280SR2, DCS-7280CR2, and 7500R2-36CQ linecard, the SandTunnel agent can restart unexpectedly with the signature "interrupt reg timeout". (193426) • On the DCS-7500 series, static host routes may become incorrectly programmed in hardware after a linecard hotswap. The workaround is to reapply the static host route configuration. (193545) • On the DCS-7280E, DCS-7500E, DCS-7280R and DCS-7500R series, using the 'tapaggregation-extended' TCAM profile may case MAC ACLs to stop working. On the DCS-7280E and DCS-7500E series, avoid using MAC ACLs for traffic steering while using the 'tap-aggregation-extended' TCAM profile. On the DCS-7280R and DCS-7500R series, avoid using 'VLAN ID' and 'Inner VLAN ID' qualifiers in MAC ACLs while using the 'tap-aggregation-extended' TCAM profile. (193743) Arista Networks Confidential 14

• If 'no snmp-server enable traps bgp' or 'no snmp-server enable traps ospf' is configured then the ribd process memory usage will begin to increase as the snmp trap triggering events occur over time. This may eventually cause the system to become low on available memory. A workaround is to remove these configuration options. (193868) • On the 7500R-8CFPX-LC module, the Denali agent will restart continuously after linecard insertion or switch boot if the startup-config has 'phy coherent-dsp' interface-level configuration for interfaces on that module. The workaround is to remove the configuration from startup-config and run 'configure replace startup-config'. (193956) • Rib agent may restart when the nexthop of a BGP learnt path which is considered to be an equal cost (ECMP) path undergoes a change such that the nexthop is recursively resolved over a route pointing to NUll0 and the nexhop was previously resolved over a route pointing to a nexthop-group adjacency or a IP(v6) route. (194520) • On the DCS-7280R and DCS-7500R series, if a speed change occurs on an interface which is configured as a mirror source the SandFap agent may restart. (194564) • On the DCS-7050X series, if PFC Watchdog with forced recovery and nondisruptive priorities is configured, a PFC watchdog timeout would occur on a transmit queue affected by a PFC pause storm immediately after a forced recovery event, regardless of the timeout interval configured. (194753) • When Port-channel interface has min-links configured, if LACP negotiation stalls on a member interface due to mismatched port-ids in peer's LACP PDUs, that member interface can get stuck in a potential aggregateable state forever. This will prevent min-links check from kicking in and shutting down the Port-channel even though active members is less than configured min-links. (194808) • On the DCS-7010, DCS-7050, DCS-7050X, DCS-7050X2, DCS-7060X, DCS-7060X2, DCS-7250X, DCS-7260X and on the DCS-7300X series, the ASIC memory scan thread in Strata agent could run into a deadlock causing the agent to restart. (195137) • Rib agent may restart unexpectedly if BGP is configured with neighbor based link-bandwidth generation (neighbor link-bandwidth default ) and the routes received from the peer are already tagged with other extended community attributes like Route-Target or Site-of-origin. (195216) • ZTP fails if the DHCP server is accessible only through the switch interfaces whose default interface speed is 25GBPS or 50GBPS. (195377)


• In an MLAG setup or on a dual supervisor modular system, the Stp agent on MLAG secondary or Standby supervisor may restart multiple times, if MLAG state changes to primary/inactive/disabled or on supervisor switchover during STP churn. This will result in traffic disruption due to Stp agent unstable restart. (195469) • On the DCS-7280E, DCS-7280R, DCS-7500E and DCS-7500R systems, vn/br tag stripping in Tap Aggregation mode will not work correctly for packets with a vlan tag after a SandFap agent restart and/or a reload. The workaround is to unconfigure and then reconfigure vn/br tag stripping. (195494) • On the DCS-7280E, DCS-7280R, DCS-7500E, and DCS-7500R series, Tap ports in Tap Aggregation mode configured with non-default ingress VLAN membership filtering and with br/vn tag stripping enabled globally will not work correctly after a system reload. The workaround is to unconfigure and then reconfigure br/vn tag stripping. (195794) • On the DCS-7500 system with DCS-7500E and DCS-7500R linecards, broadcast traffic ingressing on a DCS-7500R linecard may not get processed by cpu. This may affect ARP functionality. (195881) • The standby supervisor in RPR mode stalls and never successfully come up if 'session peer-supervisor' is run immediately after 'write mem' on the active supervisor. (196632) • On 7500R-8CFPX-LC modules, interfaces may not come up after changing modulation to 8QAM. The workaround is to change modulation to DP-QPSK or 16QAM and then back to 8QAM. (196849) • When BGP "neighbor maximum-routes" configuration is modified in a manner where that both the limit and action (warning-only) are changed, affected peers may be incorrectly idled. The problem can be resolved by administratively reseting any affected peers. (197030) • A member of an LACP port-channel may transition into 'collecting' state in the protocol without reflecting this transition to the data plane. This can result in a condition where the neighbor can start using this member to forward traffic, which will then get blackholed on reception. This issue can be identified by inspecting the output of 'show lacp internal detailed all-ports' and identifying members that have reached the 'collecting' state, but not the 'collecting/distributing' state. The workaround is to Arista Networks Confidential 16

manually flap the member interface to allow the protocol to renegotiate. (197365) • On the DCS-7050X, DCS-7060X, DCS-7250X, DCS-7260X series and DCS-7300 series, if MLAG peer routing is enabled along with VXLAN routing, an ICMPv6 Neighbor Advertisement packet received over VXLAN tunnel with source MAC matching the MLAG peer's MAC address can result in rebooting the switch with a kernel panic. (197428) • On the DCS-7250X and the DCS-7300X series switches, in an MLAG setup where fast MAC redirection is active, a packet that gets decapsulated and routed into an MLAG Port-Channel will be dropped if the local MLAG port is shut down. A workaround would be to unshut the port or to remove the port from MLAG and re-configure it. (197433) • If the next hop for a route is a neighbor on a tunnel interface, and a static ARP entry to that neighbor is deleted, the system may reset. (197536) • Enabling IGMP Snooping proxy using 'ip igmp snooping proxy' command without configuring an L2 querier using 'ip igmp snooping querier' command will result in IGMP queries not being flooded in the VLAN. A work around is to enable the querier using 'ip igmp snooping querier' to operate in IGMP Snooping proxy mode or disable IGMP snooping proxy on certain VLANS using 'no ip igmp snooping vlan proxy' (197732) • Third party (for instance bird daemon) IPv6 routes are not learnt correctly by EOS. (198097) • On the DCS-7500 series, inserting a new linecard can result in transient drops of multicast or flooded traffic, until the new linecard comes online. (198637) • Inserting a line card with a member link that is configured as part of a LAG may result in the functionality over that LAG being lost for approximately 1 minute. (198852) • On all systems, some third-party 1000BASE-T SFP+ transceivers may be incorrectly recognized as a 25GBASE AOC and the port will fail to link up. This occurs when the module EEPROM contains an erroneous extended compliance code claiming 25GBASE AOC compatibility. (198948) • The Rib agent will restart repeatedly if VRF names contain characters from "[]{}:". (198993) • On the DCS-7500E, DCS-7500R, DCS-7280E and DCS-7280R series, a switch is not able to send packets to the CPU while in Tap Aggregation mode. The Arista Networks Confidential 17

symptoms are: sFlow samples are not generated when sFlow is enabled; packets mirrored to CPU will not show up. (199409) • On the DCS-7020R series, DCS-7280R series, and DCS-7500R linecards, the SandFap agent may restart unexpectedly when a link goes down just after autonegotiation completes but before the agent can read the partner's autonegotiation advertisements. This may cause all links on the system or linecard to go down momentarily. This affects autonegotation at 1G speeds with 1000BASE-SX or 1000BASE-LX transceivers and 25G/40G/50G/100G speeds on QSFP+/QSFP28 ports. A workaround is to force speeds on those interfaces using the "speed forced 1gfull|25gfull|40gfull|50gfull|100gfull" CLI command in interface configuration mode. (199473) • The Rib agent can restart unexpectedly when OSPFv3 is translating Type 5 LSAs to Type 7 LSAs if the LSAs do not have external route tags. Ensure all Type 5 LSAs which must be translated have route tags. (199919) • The Rib agent may restart unexpectedly when running OSPFv3 with multiple NSSA areas and redistributed routes. (199920) • When "soft-reconfiguration inbound all" is enabled for a BGP peer and if a path update with a AS loop is received for the BGP best path which is also the head ECMP path, then the path is made ineligible and forwarding information is updated based on the new best path for the destination prefix. If a subsequent path update is received for the previously ineligible path now with a valid AS-PATH without an AS loop and if this path is selected as the best path again but this time without ECMP, then BGP may fail to recompute the forwarding information; it will carry over stale adjacency from before the path became ineligible due to AS loop. The corresponding FIB route entry would then point to a FEC with multiple nexthops (ECMP) when ECMP is not expected. A workdaround would be to disable "soft-reconfiguration inbound all" for the BGP peer and resetting the BGP session for that peer. (200050) • On switch models other than the 7160 series, systems using a large number of nexthop groups may reset in uncommon circumstances. (200135) • On the DCS-7280E, DCS-7280R, DCS-7500E, and DCS-7500R series, if a source mirror is removed while it is down due to speed change it will remain a source mirror when it is brought back up. To fix this issue if encountered, reconfigure and unconfigure the source mirror while the interface is up. (200388)


• The Rib agent may restart unexpectedly if BGP has received more than 256 paths for an ECMP prefix when the platform max-ecmp limit is 128. (200389) • On the DCS-7280TR-48C6 and DCS-7280TRA-48C6, interfaces which are configured to be less than 10G when the SandFap agent starts up will not be able to achieve linerate if they are configured to be 10G at a later time. (200804) • If multi-agent mode is configured for routing protocols on a BGP router in a network with route-reflectors, the Bgp agent on the router may consume a lot of CPU and memory processing received BGP updates. (201182) • On the DCS-7500E, DCS-7500R, DCS-7280E and DCS-7280R series, a small number of source packets will not be mirrored from a monitor session. The probability is approximately 1 in 65k source packets with the E series, and 1 in 15M with the R series. (201305) • When an as-path access-list configuration is removed and quickly added or modified via the config replace feature, the Rib agent may restart. (202363) • If the IS-IS router receives a purge LSP within a small window of time between the expiry of that LSP and its removal from the LSP database, then there is a possibility of an inconsistent state in the LSP database resulting in a Rib agent crash. (202390) • On DCS-7280QR-C72, DOM sensor values on ports 7 - 30, 43 - 66 (all inclusive) are always reported as 0 in SNMP queries. (202421) • On the DCS-7500R2 and DCS-7280R2 series, SandTunnel agent restart or parity error correction in the EDB block of the forwarding ASICs may cause packets to be misdirected if the following features were configured: VLAN Translation, Dot1q Tunnel, Tap Aggregation, and Sub-interfaces. The workaround is to de-configure and then re-configure these features. (202434) • On the DCS-7250QX and DCS-7300X series, with the VXLAN routing feature enabled, if both overlay and underlay routes are reachable over ECMP nexthops, the StrataL3 agent may restart unexpectedly. (202509) • In a given BGP update-group (BGP rib out), if a peer gets stalled (is not able to write because of filling up the buffer) at the time of sending the last withdrawal or the last route update, pending queued announcements may not be sent to all the peers in the update group. (203772) • On 7500E-36Q-LC linecards, changing port speed configuration from "speed auto 40gfull" to "speed forced 10gfull" on ports 3, 9, 15, 21, 27, or 33 will cause a forwarding agent restart and all links on the card will go down during the restart. Arista Networks Confidential 19

The workaround is to configure "speed forced 40gfull", then configure "speed forced 10gfull", on those ports. (204505) • The EOS Extra package "IgmpSnoopingSrv.swix", providing an IGMP Snooping diagnostic server, does not work. (204732) • On DCS-7050X, DCS-7250X, DCS-7060X, DCS-7260X, DCS-7300 series switches, forming an MLAG switch pair( S1 and S2), if the MLAG fast MAC redirect feature is activated on either switch S1 or S2, for an MLAG which is being used as egress interface for a core SVI, will cause VXLAN encapsulated packets to be dropped on the switch (S1 or S2) where the fast MAC redirect is activated. The work around is to : - To ensure that MGLAG is not used an egress interface for core routes. - To change peerlink to be a single interface ( which will deactivate fast mac redirect ). (204885) • Link-flap might cause inconsistent ARP entries in the kernel and other databases until the ARP is flushed from the kernel or relearned. (205280) • On DCS-7160 series switches, during VRRP active/backup role determination among the VRRP enabled peers, there is a transient race condition that might cause the XpMact agent to restart when programming VRRP group MAC address. The system is expected to recover after XpMact agent restarts. (205415) • On the DCS-7280E, DCS-7280R, and DCS-7500R series switches, after power cycling a line card, some ARP entries are not programmed in hardware. (205444) • On the DCS-7280E, DCS-7280R, and DCS-7500R series switches, when VXLAN host routes are programmed in LPM, some ARP entries are not programmed in hardware after switch reload. (205616) • In an MLAG setup, 'spanning-tree portchannel guard misconfig' may incorrectly errdisable MLAG port-channels when PVST BPDUs are being hashed to primary and secondary port-channel members. Remove 'spanning-tree portchannel guard misconfig' configuration in spanning-tree PVST mode. (205849) • For the 7020TR-48 and 7020TRA-48 products, when interfaces ethernet49 to ethernet54 are configured with 1G speed, outgoing packets with size greater than or equal to 1600 bytes and less than 9214 bytes will cause these ports to get stuck. (206078)


• On the DCS-7500R2 series and the DCS-7280R2, MAC is not learnt with incoming IPv6 packets on SVIs (206163) • When a BGP ECMP path that is redistributed into a IGP protocol shrinks due to a peering link going down, the Rib agent may restart. (206217) • On the DCS-7500E, DCS-7500R, DCS-7280E and DCS-7280R series, if a port with QoS ACL policing is configured as a source interface in a monitor session with a filtering ACL, then QoS ACL policing on this port will not work. Remove the affected port from the monitoring session or remove the filtering ACL from the monitoring session will restore QoS ACL policing. (206515) • On the 7050X2 series, shutting a port with 1000BaseX/T SFP in it can result in continued port flaps. (207107) • On all platforms with QSFP+, QSFP100, MXP, CFP2 or CFP-X ports, optical transceivers may overheat. Cooling fans will not be running at 100% despite the overheat condition. Workaround is to manually override cooling fan speed using the command: environment fan-speed override The default recommended settings are 80 for reverse airflow systems and 60 for forward airflow systems. These settings may need to be adjusted depending on inlet temperatures and the power of the optics used. (207171) • On DCS-7050X, DCS-7060X, DCS-7250X, DCS-7260X and the DCS-7300X series, it is not possible to configure the Port-Channel load balance hash algorithm for IP packets to include only the TTL field. The CLI configuration command "port-channel load-balance trident fields ip ttl", will cause the switch to include source-ip, destination-port, source-port and destination-port in the hash beside TTL field. Similar issue applies to IPV6 packet hashing configured via the "port-channel load-balance trident fields ipv6 hoplimit" command. (209926) • The Rib agent may unexpectedly restart when it sends a BGP Update to a Route-reflector Client peer if this BGP update contains an unexpected Originator-ID originally received from an EBGP peer. (210512) • On DCS-7300 and DCS-7500 series switches, when a linecard is replaced by a different model the interface configuration for that linecard is intentionally deleted in the running-config and the new model linecard SKU is saved in /persist/sys/Linecard.cfg. However, if the configuration has description containing 'Ethernet' then the configuration cleanup will not be completed. The new card model name configuration is not saved in /persist/sys/Linecard.cfg file and the system will delete interface configuration in future reloads or card hotswaps. Arista Networks Confidential 21

To workaround the problem, the /persist/sys/Linecard.cfg file for affected slots must be deleted and the affected cards hotswapped using hidden CLI commands 'platform module Linecard '. The new interface configuration will have to applied after that. (211179) • On the DCS-7050X, DCS-7250X, DCS-7300X series, a parity event on FP_GLOBAL_MASK_TCAM may cause a slice agent to restart because of Hard Fault. This may disrupt traffic flow temporarily. (214590) • When OSPF is configured on a port-channel interface with OSPF interface area configuration and there are multiple areas configured in the OSPF instance and a port-channel member interface flaps, the OSPF port-channel can be removed from the OSPF network. Removing the interface from the OSPF network can cause OSPF adjacency disconnection and traffic loss. To work around the problem, OSPF global area configuration should be used to add the port-channel to the OSPF network instead of OSPF interface area configuration. (214599) • When OSPF expires a non-self-originated external or summary LSA, and that LSA announces a prefix that is less specific than a self-originated LSA with the same type and LSID, OSPF can unsuppress the conflicting self00 80 99 t. originated /32 LSA when it shouldnâ 00 If there is a self-originated LSA for the same less specific prefix and type, then the resulting state can cause the Rib agent to consume 100% of the CPU as well as restart unexpectedly. It can also cause OSPF adjacencies to get stuck in the exchange state. (214914) • A software bug in the system can cause all processes to restart on the switch. (216432) • When the /var/log file system is full, new SSH connections might hang until files have been deleted on /var/log (219202)

General Known Software Caveats in 4.19.1F • The Line and Fabric cards may be lost after release upgrade. As result SSO switchover, will be disabled. Additional forcefull switch restart (and traffic disruption) shall fix the issue. (184007) • On the DCS-7500RA and DCS-7280RA switches, applying an ACL that exhausts the hardware resources will result in the traffic drop on other interfaces on the switch that have ACLs applied. (213600)


Layer3 Known Software Caveats in 4.19.1F • On the DCS-7010T, when PBR is configured, Strata agent can restart when there is a port channel member flap exactly at the same time as PBR configuration is changed. (135734) • When IpRib agent restarts and at the same time ARP entry for a nexthop is learned, then IpRib agent could crash again and again repeatedly while starting. (204997) • OSPF and OSPFv3 do not consider point-to-point interfaces when selecting a forwarding address for a NSSA Type 7 LSA. There is a workaround for the issue. In order to have an interface considered when selecting a forwarding address for a NSSA Type 7 LSA, the OSPF or OSPFv3 point-to-point inteface configuration can be removed. (212542) • OSPF and OSPFv3 do not consider point-to-point interfaces when selecting a forwarding address for an AS External Type 5 LSA. There is a workaround for the issue. In order to have an interface considered when selecting a forwarding address for an AS External Type 5 LSA, the OSPF or OSPFv3 pointto-point inteface configuration can be removed. (212594)

DCS-7160 Known Software Caveats in 4.19.1F • On the DCS-7160, DCS-7280RA and DCS-7500RA series switches with AlgoMatch ACLs, the ACL must not contain any rule that matches on IP address/mask where the IP address/mask is not a prefix. If such a rule is present, that rule and all subsequent rules in the ACL will not be programmed in the hardware. (217795)

DCS-7500 Known Software Caveats in 4.19.1F • On the DCS-7500E, DCS-7500R, DCS-7500R2, DCS7280E, DCS-7280R and DCS-7280R2 series while in Tap Aggregation mode, if a packet matches two separate class maps with only the lower priority one having a destination with truncation enabled, then this packet will be forwarded to both the regular destination of the higher priority class, and the truncation destination of the lower priority class. Always configure a tool destination with truncation as part of the higher priority class. (196875) • On the DCS-7500E, DCS-7500R, DCS-7500R2, DCS-7280E, DCS-7280R and DCS-7280R2 series while in Tap Aggregation mode, if a traffic steering policy contains both a deny rule and a destination with egress truncation, then the denied traffic will be forwarded erroenously to the destination Arista Networks Confidential 23

with egress truncation. Configure deny rules as a separate class-map in a traffic steering policy and set the destination to an unused tool port. (203992) • On the DCS-7500 and DCS-7280 series, MPLS tunneling of ICMP TTL Exceeded messages is enabled by default. Shutting down the MplsUtilLsp agent will prevent the MPLS hops from being visible in IP traceroute but will disable MPLS ping and MPLS traceroute features. (205854) • On the DCS-7280E, DCS-7280R, DCS-7500E, and DCS-7500R series switches, at high packet rates, the DSCP value in the outer IP header of VXLAN encapsulated packets egressing from the switch might be incorrect. (206170) • On the DCS-7280E, DCS-7500E, DCS-7280R, DCS-7500R, DCS-7280R2 and DCS-7500R2 series while running in Tap Aggregation mode with the tapaggregation-extended TCAM profile configured, MAC access lists used in traffic steering policy maps may stop functioning following a system restart. Unconfigure the "extended" TCAM profile, and reconfigure it again. (213177) • On the DCS-7280, DCS-7500E, and DCS-7500R systems, VXLAN traffic that is decapsulated and bridged will incorrectly apply any IP Router-based ACLs from the destination VLAN. (213755) • On the DCS-7500R modular switches with 7500RA linecards and on DCS-7280RA switches, when an IPv4 or IPv6 egress acl is applied on a lag interface, the egress traffic out of the port-channel will all go out of one of the members. (214486)

Multicast Known Software Caveats in 4.19.1F • When there is a lot of traffic to the CPU, the router might drop PIM hellos and lose neighborship. (215660)

DCS-7300 Known Software Caveats in 4.19.1F • On DCS-7050TX/SX/QX and DCS-7260QX series switches, on non default VRF's, PIM Bidir packets hitting fastdrop route get incorrectly punted to CPU. (211442) • On the DCS-7050X, DCS-7250X, and DCS-7300X series, a single-event upset (SEU) in the MMU_ENQ_PBI_DB table may lead to packet loss. This condition


will persist until it is manually cleared. A forwarding agent restart is required to clear the error and forward traffic normally. (213017)

General Limitations And Restrictions in 4.19.1F • On 7500R2M-36CQ-LC linecards, frames less than 64 bytes are dropped (as expected), but not counted. Command "hardware phy cmx42550 diag read64 mac line 0x138" can be used to dump the counter for the purpose of debugging". (218429)

DCS-7300 Limitations And Restrictions in 4.19.1F • On the DCS-7250X, DCS-7260CX and DCS-7300X series, if IP access-list or IPv6 access-list is configured on SVI in egress direction, then permit rules in the access-list may count the packet the twice. Double accounting for the permitted packet will occur only if the packet ingresses and egresses out of different chips. (177858)

Conditions and Impacts The release notes listed above use shorthand terminology to refer to conditions that arise frequently in connection with bugs. This section explains each condition and its impact in more detail. Condition: Rib agent restart Impact: When the Rib agent goes down, all routing sessions are terminated; all BGP sessions drop, all OSPF adjacencies are lost, and neighbors stop forwarding traffic to us. When the Rib agent restarts, adjacencies are re-formed, and, after protocol convergence, traffic flows through us again. In most cases where there is adequate networt-level redundancy, application-visible impact should be minimal Condition: Rib agent hang Impact: When the Rib agent hangs, routing protocols stop running. Routing peers will generally time out their session with the hung Rib agent, withdrawing routes accordingly. Meanwhile, the device continues to forward packets based on existing routing state. During this time, the device will not respond to routing topology changes. After a heartbeat timer expires (typically 10 minutes), the Rib agent restarts. In networks with sufficient redundancy, application impact of a Rib agent hang is usually low, because there is no data path disruption. However, in conjunction with other network changes (such as link failure or introduction), routing loops may occur. Condition: kernel crash Impact: A kernel crash has impact similar to issuing the "reload now" command. Arista Networks Confidential 25

All links go down and all forwarding stops. Then, the system reloads, which may take up to 15 minutes. In a network with adequate redundancy, there should be minimal traffic loss associated with this period. After reload, links come up and protocols (LACP, STP, BGP, etc) reconverge. Protocol reconvergence is not necessarily hitless, depending on topology and configuration. For example, a switch may advertise a prefix to a connected subnet before STP has converged. However, any associated outage should be short lived, typically lasting around 30 seconds. The MLAG-SSO feature can dramatically reduce the window of disruption. Condition: forwarding agent restart Impact: A forwarding agent restart typically results in the switch ASIC being reset. It clears packet memory, dropping all packets currently in the switch, and causes all links to go down. The restart can take up to 45 seconds, during which time all links are down. Once the restart completes, all links come back up automatically, followed by protocol reconvergence (MLAG, LACP, STP, BGP/OSPF/ISIS, etc). Depending on which protocols and options are in use, the reconvergence may take a few seconds up through several minutes. On modular switches, the impact of forwarding agent restart is limited to the affected line card; that is, if the forwarding agent for line card 3 restarts, then all ports on line card 3 bounce and protocols on those ports reconverge, but ports on other line cards are unaffected.


Arista Eos Rn-4.19.1f

Recommend Documents