These notes summarizes PICOS 4.1 new features, new hardware, known bugs, and bug fixes. Best practices recommend that you read all the content before upgrading to this release. For more detailed feature information, refer to the configuration guides.
Layer 2 and Layer 3
DHCPv6 relay is the IPv6 equivalent of the DHCP relay for IPv4. The DHCPv6 relay agent passes DHCPv6 solicitation or request to the DHCPv6 server on a different subnet and get the returned DHCPv6 advertise or reply from the DHCPv6 server back to the DHCPv6 client.
Support OSPFv3 for IPv6
OSPFv3 (Open Shortest Path First version 3) is added to support IPv6. Please note that VRF over OSPFv3 is not available in this version (4.1.0).
Multiple OSPF Instances
Eight OSPF processes are started if multi-instance is enabled. Each OSPF instance, identified by an unique number (1 - 8), can be configured separately. Please note that multiple OSPF instances are supported only in the default VRF with OSPFv2.
Support VRRPv3 for IPv6
VRRPv3, with IPv6 introduced, can bundle 2 physical PICOS routers as one single virtual router. The ND (Neighbor Discovery) protocol is enabled on the virtual IPv6 VRRP interfaces by default. In particular, the RA (Router Advertisement) packets will be sent out from the VRRP virtual interfaces periodically.
Update ping and traceroute for IPv6
With upgrade to Debian 10.9, CLI commands, ping and traceroute, are updated to support IPv6. And ping6 and traceroute6, are removed from CLI under operational mode.
Added the support of multi-line banners with the newly-added CLI command "set system login multiline-banner ...".
VRF Route Leaking
Routes can be leaked from one VRF to a different VRF. In version 4.1.1, we support VRF route leaking in case of static routing.
IPv6 Support in ACL
Support IPv6 in ACL rules including firewall filter and NAC downloadable/dynamic ACL rules. IPv4 and IPv4 ACL rules can be configured in parallel. If a packet matches both a IPv4 rule and a IPv6 rule, the IPv6 rule has higher priority.
Specify Source Interface for SNMP Trap Configuration
Support to source interface for SNMP Traps. The source interface could be loopback or l3-interface. The specified source interface is used to derive the source IP address for the SNMP traps sent, so that traps received from each switch will always have a single consistent source IP address.
Loopback Inband Interface
IP addresses configured on loopback interface in default VRF namely lo can be used for inband management.
The content of login annoucement with multiple lines can be configured by new added CLI command "set system login multiline-announcement ...".
LACP is a mechanism to keep a link available on a LACP LAG port even though the LACP is not ready on the peer device such as the case of PXE (Preboot eXecution Environment) client-server. In particular, LACP fallback can be enabled on MLAG port.
MSDP Support and Anycast RP
MSDP, Multicast Source Discovery Protocol, is used for sharing active multicasting sources between different RPs. Anycast RP is a specific application scenario of MSDP, which can provide load sharing and redundancy in PIM-SM networks. In case of anycast RP, multiple RPs are configured with the same IP address on loopback interfaces. A source or receiver will select the closest RP based on the routing distance.
Timestamp on NAC Session
For a NAC session which is authorized for a specific port (802.1X) or client (MAB), when execute command "run show dot1x interface gigabit-ethernet xxxx", display the timestamp indicating that when the session is ready.
Add a new VXLAN decapsulate-mode
Add a new decapsulate-mode "service-vlan-per-port". The decapsulated packet can be tagged or untagged dynamically based on the setting on the output port.
Anycast Gateway for EVPN Distributed Architecture
The anycast gateway can be configured on VTEPs for each attached subnet in case of EVPN distributed architecture including symmetric mode and asymmetric mode. With anycast gateway enabled, a host can use its local VTEP as its default gateway to send traffic out of its IP subnet. Additionally, one single IP address can be configured on all hosts within an VNI.
ARP/ND suppression enables VTEPs to prevent ARP/ND from flooding over VXLAN tunnels. In case of EVPN deployment, IP/MAC pairs can be exchanged between VXLAN VTEPs via EVPN type 2 messages. Therefore, it's not necessary that ARP/ND packets go through VXALN tunnels to have or update the MAC addresses of the hosts in remote site.
Support 31-bit (IPv4) or 127-bit (IPv6) Subnet Mask
Support to configure IPv4 addresses with 31-bit subnet mask as well as IPv6 addresses with 127-bit subnet mask.
Include VRF in tech_support
The information of routing and ARP and ND (Neighbor Discovery) with respect to configured VRF instances is included in tech_support report.
Enhance Site License to use Dell Service Tag ID
Service tags on Dell platforms are supported in site license key in addtion to serial numbers.
Enable FEC on 25G Ports
Support FEC (Forward Error Correction) on 25G ports under both OpenFlow and L2/L3 mode. Please note that CL74 (Base-R) instead of RS-FEC is supported for 25G ports on Tomahawk/Tomahawk+ platforms such as AS7312 and AG5648.
AS6812_32X is available in 2.11.x. Support AS6812_32X in 4.1.2.
Add AG7648 in 4.x.x
Support AG7648 in release 4.1.3.
Multi-Actions under Crossflow Mode
If no flow hit by the incoming packet, it will be trapped to CPU and sent to the Controller. With multi-actions flow such as "in_port=5,actions=normal,controller", the packets received on port 5 will be sent to controller as well as be forwarded out normally.
Migrate NPB to 4.1.1
Migrate NPB application to 4.1.1 from 3.2.x.
Layer 2 and Layer 3
DNS name-server and DNS Search List
If users configure DNS name-server or DNS search list from PICOS CLI, the associate configuration included in /etc/resolv.conf should not be updated from DHCP server.
Cannot Access to SNMP Agent
If users enable management VRF and Inband connection as following CLI commands, SNMP agent stopped responding.
The Size of NETCONF.events Increases without Limit
If users configure NETCONF on the switch, the size of NETCONF log file /tmp/stream/NETCONF.events will increase without limit. Add a mechanism of rotation to fix this issue. Keep the size of this file within 2M.
Crash if Users Configure port-mode via NETCONF
The NETCONF process might crash if configure port-mode to "trunk" and then back to "access" repeatedly via NETCONF.
Multicast Routes in VRF not Applied to ASIC
The multicast routes specific to a VRF cannot be applied to the hardware ASIC. Therefore, the multicast traffic within this VRF is pumped up to CPU and routed out by the multicast routes in the kernel. This issue was fixed in 184.108.40.206. Additionally, as a known issue of FRR, if configure join-prune-interval or keep-alive-timer or register-suppress-time, they will be duplicated in the configuration both under default VRF and other VRFs even if PIM is used there.
Multicast Packet Drop Caused by Aging-out of Multicast Routes in the Kernel
Generally, the multicast traffic hits the multicast routes in ASIC and go through the data plane. The corresponding multicast routes in the kernel will be aged out if no traffic hit within a specific timeout. The multicast routes in the kernel should be updated based on the status of the multicast routes in ASIC.
Upgrade2 from PICOS 3.x.x to PICOS 4.x.x
The configuration in 3.x.x will not be brought into 4.x.x potentially. Instead the configuration /pica/config-4.x/pica_startup.boot will be loaded when update to 4.x.x. In case of upgrade2, the old version will be kept in the switch in order for returning to it latter. Additionally, the 3.x.x configuration will be copied to /udata/pica/config-3.x/picos_startup.boot. The platforms with SquashFS including N3000 and N3100 don't support this feature.
NETCONF is updated to support 4.x.x changed functionalities including static routing, VRRP VXLAN and L3 interface. The Yang model support for BGP and OSPFv2 & OSPFv3 are not added in 4.1.1.
Encrypt Passwords Configured in BGP & OSPF
Encrypt the passwords appearing in BGP & OSPF set commands.
Include FRR Config in the tech_support
By including FRR config in the tech_support which would be sent back from customer for trouble shooting, it is much helpful to identify configuration consistency issues between PICOS CLI and FRR runtime configuration.
It's possible that SNMPwalk fails when switch has multiple VLANs enabled inband because the returned SNMP reply message can only be sent back via the L3 interface on which the associate SNMP query message is received. It is fixed in 4.1.1 by lookup the L3 routing tables to send the SNMP reply messages to SNMPwalk client.
Image File is Not Removed after Upgrade
In case of upgrade2, if place the image file of new version in /udata instead of /cftmp, this image file will not be removed when update to new version.
TACACS+ User in PICOS CLI Prompt
If a TACACS+ user login to the switch, the TACACS+ user name should be showed in PICOS CLI prompt in stead of the associate local user names such as admin or operator.
Keep DNS Servers in Configuration Order
The nameservers stored in resolv.conf should be in the order we configure "system dns-server-ip" in the switch.
Suppress VRRP rsyslog Messages
It does not make sense to print the same rsyslog messages "Cannot find IF vlanXXXX" repeately if there are 2 VRRP groups configured on the same subnet.
The VXLAN Traffic cannot be Terminated
The VXLAN packet with a VRRP virtual IP address as the underlay external destination IP address cannot be terminated. This issue is fixed in 4.1.1.
Suppress False FRR Error Messages
When configure OSPF, it may print below false rsyslog message.
802.1X Authorization Failure on VXLAN Access Port
When reset VXLAN configuration from AmpCon SDN controller, 802.1X authorization on the VXLAN access port may be failed. This issue is fixed in 4.1.2.
Different MAC Addresses of VRRP IP Returned
In case of active-active VRRP over MLAG, when a client requests the MAC address binding to VRRP virtual IP address via ARP/NS, different MAC address may be returned. This issue is fixed in 4.1.2.
Fixed IPv6 Routing Errors
For specific route entry in hardware, if the next hop is changed to an immediately connected host or re-directed to a l3 interface, this route entry may be removed mistakenly. This issue was fixed in 220.127.116.11. Additionally, IPv6 NA (Neighbor Advertisement) packets should not be routed out of the switch. If a NA is pumped to CPU, will check its source IP address against the ingress L3 interface and make sure the associate IPv6 neighbor is added to the right IPv6 subnet.
Remove MTU Restriction on VXLAN Traffic
The restriction that packet with size greater than 1422 bytes cannot go through VXLAN tunnel is removed in 18.104.22.168.
Access Port Disappear After Add/Delete VXLAN Repeatedly
Repeat the procedure, delete VXLAN configuration and then rollback, access ports in a VXLAN instance may disappear when "run show vxlan vni xxxx".
Login ACL Rules Should not Applied to SNMP Access
The configured login ACL rules are also applied to SNMP access. This issue was fixed in 4.1.3.
Cannot Rollback to Default Configuration
By integrating FRR code since 4.0.0, it cannot roll back to default configuration with CLI command "rollback default" under configure mode. This issue was fixed in 4.1.3.
With multiple TACACS+ servers configured, if one TACACS+ server is not reachable, sshd may crash with "double free or corruption" printed on the console terminal. This issue was fixed in 4.1.3.
Fail to Install Site License on Dell Platforms
It fails to install site license on Dell platforms with serial numbers having 20 characters. This issue was fixed in 4.1.3.
Don't Display Output of ntptime in NTP Status
When display NTP status with CLI command "run show system ntp-status", will not call ntptime which is used to display or set the kernel variables used by NTP. And the confusing error messages of ntptime, which don't make sense for users, will not be displayed.
Have VXLAN Traffic Loop if Underlay Routing Being Changed
If underlay routing is being changed constantly, a traffic loop of VXLAN overlay may appear. Namely, on a specific VTEP, a MAC address of a host is learned on both access port and the network port with regarding a VXLAN instance. This issue was fixed in 4.1.3.
Multiple Output Tunnel Ports in a VXLAN Flow
Multiple tunnel ports can be added as the output ports in a VXLAN flow. It requires that the VXLAN ports must have the same VNI (Virtual Network Identifier) on the ingress side.