* [ACL] enable ACL FC when genereting config from minigraph but disable by default
Why I did it
To support ACL counters on Flex Counter Infrastructure.
How I did it
Enable ACL FC in init_cfg and minigraph. Disable when genereting configuration from preset.
How to verify it
Together with depends PRs. Run ACL/Everflow test suite.
Signed-off-by: Stepan Blyshchak <stepanb@nvidia.com>
Added get_system_mac support for cisco-8000 device
Why I did it
This is required to assign unique MACs to namespaces on Backend and Frontend asics
for cisco-8000 platforms
How I did it
Add vendor specific hook in get_system_mac. The MAC address are read from /profile.ini
file
How to verify it
config load_minigraph on a multi-asic system
config save
check mac addresses in DEVICE_METADATA section in each config_db json file
What I did:
Fix the typo in Internal Peer Group template for Packet-based Chassis.
Address Review comments of PR: [chassis-packet] minigraph parsing and BGP template changes #8966
- Static Route Parsing for Host
- Formatting of chassis port_config.ini
Signed-off-by: Neetha John <nejo@microsoft.com>
Why I did it
Storage T0's have all vlan members as tagged
How I did it
Since currently minigraph does not have a unique way to identify if a vlan member is tagged/untagged and to ensure other scenarios are not broken, the logic used is to just update the vlan member type as 'tagged' when we determine that it is a storage backend device. This change will apply only to storage backend T0's since storage backend T1's will not have vlan member information
How to verify it
Updated the storage backend T0 testcases to check for tagged vlan members
Added testcase to check if a T1 and backend T1 device generates an empty vlan member table
Existing vlan member testcases are good enough for checking if any regression has been caused for regular T0's
Build sonic_config_engine-1.0-py3-none-any.whl successfully
Fix the check used to wait for interfaces to come up. The group name in
the supervisor config files has changed from isc-dhcp-relay to
dhcp-relay.
Also, in the wait script, wait 10 additional seconds after the vlans,
port channels, and any interfaces are up. This is because dhcrelay
listens on all interfaces (in addition to port channels and vlans), and
to ensure that it stays in a clean state during runtime, wait some extra
time to make sure that those interfaces are created as well.
Signed-off-by: Saikrishna Arcot <sarcot@microsoft.com>
1. Changes for Generation LC-Graph for packet-based chassis.
2. Added Support Ipv6 Peering on Loopback4096 for voq also
3. Updated asic topology yml files to be offset of slot
4. Made slot_num to take string slot<number> instead of number
5. Consolidated template_dpg_voq_asic.j2 into dpg_asic.j2
6. Remove Loopback4096 from asic topology and parse as dut invertory for
multi-asic
7. Updated topo_facts parsing for asic topology_
8. Internal BGP Session rename from <VoqChassisInternal> to <ChassisInternal> and take switch_type as value.
Signed-off-by: Abhishek Dosi <abdosi@microsoft.com>
Depends on Azure/sonic-utilities#1626
Depends on Azure/sonic-swss#1754
QOS tables in config db used ABNF format i.e "[TABLE_NAME|name] to refer fieldvalue to other qos tables.
Example:
Config DB:
"Ethernet92|3": {
"scheduler": "[SCHEDULER|scheduler.1]",
"wred_profile": "[WRED_PROFILE|AZURE_LOSSLESS]"
},
"Ethernet0|0": {
"profile": "[BUFFER_PROFILE|ingress_lossy_profile]"
},
"Ethernet0": {
"dscp_to_tc_map": "[DSCP_TO_TC_MAP|AZURE]",
"pfc_enable": "3,4",
"pfc_to_queue_map": "[MAP_PFC_PRIORITY_TO_QUEUE|AZURE]",
"tc_to_pg_map": "[TC_TO_PRIORITY_GROUP_MAP|AZURE]",
"tc_to_queue_map": "[TC_TO_QUEUE_MAP|AZURE]"
},
This format is not consistent with other DB schema followed in sonic.
And also this reference in DB is not required, This is taken care by YANG "leafref".
Removed this format from all platform files to consistent with other sonic db schema.
Example:
"Ethernet92|3": {
"scheduler": "scheduler.1",
"wred_profile": "AZURE_LOSSLESS"
},
Dependent pull requests:
#7752 - To modify platfrom files
#7281 - Yang model
Azure/sonic-utilities#1626 - DB migration
Azure/sonic-swss#1754 - swss change to remove ABNF format
To Fix#8697 . The config load_minigraph initializes 'admin_status' to up when platform.json has DPB configs. This doesn't happen when using port_config.ini
The update minigraph has logic to initialize only the ports whose neighbors are defined or those belonging to portchannel
However, a change was introduced to have default admin status to be 'up' in portconfig.py when the minigraph was using platform.json
This will lead to sanity check failure in sonic-mgmt and thus no test cases could be run
#### Why I did it
Allow cfggen to work on system without ports in platform.json or in port_config.ini
#### How I did it
Add json write of PORT section only if the dictionary that contains the ports is not empty.
#### How to verify it
sonic-cfggen -k ACS-MSN3700 -H -j /etc/sonic/init_cfg.json --print-data
Fix for sonic-cfggen exception during platform string read during fresh install and start of sonic in multi asic, /var/run/redisX/ is created after database docker is started.
In version 3.0.0, If a broadcast address is specified in
/etc/network/interfaces, then when ifup is run, it will fail with an
error saying `'str' object has no attribute 'packed'`. This appears to
be because it expects all attributes for an interface to be "packable"
into a compact binary representation. However, it doesn't actually
convert the broadcast address into an IPNetwork object (other addresses
are handled).
Therefore, convert the broadcast address it reads in from a str to an
IPNetwork object.
Also explicitly specify the scope of the loopback address in
/etc/network/interfaces as host scope. Otherwise, it will get added as
global scope by default. As part of this, use JSON to parse ip's output
instead of text, for robustness.
Signed-off-by: Saikrishna Arcot <sarcot@microsoft.com>
* Update default cable len to 0m for TD2 (#8298)
* Update sonic-cfggen tests with the correct cable len
Signed-off-by: Neetha John <nejo@microsoft.com>
As part of the buffer reclamation efforts for TD2, setting the default cable len to 0m which means unused ports will have a cable len of 0m.
Why I did it
To align with the changes in Azure/sonic-swss#1830
How to verify it
- With the default cable len set to 0m and the associated changes in swss, CABLE_LENGTH table had '0m' set for unused ports and accordingly more space was reserved for the shared pool
- Cfggen tests passed with the cable len update
Why I did it
In the config_db.json generated by minigraph "admin_status" attribute is missing for the VOQ inband interface port in the PORT table.
How I did it
Changes done to add admin_status attribute for voq inband interface port, if it exists in the PORT table keys.
Why I did it
There are scenarios that End-of-RIB comes from a part of the peers arrives after reconciliation. In such scenarios, if the route selection deferral timer has the default value of 360 seconds, FRR would not set up routes and all routes would be removed after reconciliation. This PR reduces the route selection deferral timer so that at least routes to parts of the peers get restored at the point of reconciliation.
Fix#7488
How I did it
Reduce route selection deferral timer for bgp graceful restart to 15 seconds.
lazy_re had an issue when importing sonic-cfggen in another application that
uses re.search(). There is no much improvement of lazy_re today after many
other good optimization work done for sonic-cfggen. It served as a quick
temporary solution.
Some quick test for fast-reboot and warm-reboot done on top of 201911 branch:
Fast-reboot: from ASIC reset to ports in up state:
with lazy_re: 18 sec
without lazy_re: 18 sec
Warm-reboot: LAG restoration time:
with lazy_re: 73 sec
without lazy_re: 72 sec
So, there is no real optimization since the number of sonic-cfggen calls is greatly
reduced in latest SONiC. This means it is time to revert this change.
Signed-off-by: Stepan Blyschak <stepanb@nvidia.com>
Why I did it
Currently SONiC use the 'isc-dhcp-relay' package to allow DHCP relay functionality on IPv4 networks only.
This will allow the IPv6 functionality along the IPv4 type.
How I did it
Edit supervisord template to start DHCPv6 instances when configured to do so on Config DB.
Align cfg unit test to the new change.
Add DHCPv6 relay minigraph parsing support and a suitable t0 topology xml file for UT.
How to verify it
Configure DHCPv6 agents as described on the feature HLD: Azure/SONiC#765
Test it with real client/server with IPv6 or use the dedicated automatic test: Azure/sonic-mgmt#3565
Signed-off-by: Shlomi Bitton <shlomibi@nvidia.com>
* Split docker-dhcp-relay.supervisord.conf.j2 template into several files for easier code maintenance
Why I did it
Allow deploying DHCPv6 servers following the implementation PR: #7772
How I did it
Add DHCPv6 to minigraph.py on sonic-cfggen tool and improve the unit test to cover this change.
How to verify it
Try to deploy a switch with DHCPv6 servers.
Signed-off-by: Shlomi Bitton <shlomibi@nvidia.com>
A recent version of contextlib2 (https://pypi.org/project/contextlib2/21.6.0/#history) has broken Python2 compatibility, so the version picked up by netaddr when using Python2 must be specified, or else builds fail
Co-authored-by: Tom Zhu <tom.zhu@metaswitch.com>
Discussion and requirement in Chassis discussion forum to NOT make the asic-id field in the DEVICE_METADATA mandatory. If this field "asic-id" is not present the orchagent will be started without the -i <asic_id> parameter
Ref: https://github.com/Azure/sonic-buildimage/blob/master/dockers/docker-orchagent/orchagent.sh#L39
How I did it
Made the check to see if the asic-id is valid and update the asic-id field in the DEVICE_METADATA
Signed-off-by: Neetha John <nejo@microsoft.com>
Why I did it
The current logic generates 'VLAN_SUB_INTERFACE' table if the device type is backend and cluster name contains 'str'. This is not a reliable method to determine a storage backend device
How I did it
Updated the logic to generate 'VLAN_SUB_INTERFACE' table if any of the following conditions hold true
- device is of type backend and ResourceType attribute is None
- device is of type backend and ResourceType attribute contains "Storage"
- device is of type backend and graph contains "Subinterface" section
Also updated the logic to set "is_storage_device" to True
- For Backend, if any of the above conditions hold true
- For Frontend, if ResourceType attribute contains "Storage"
How to verify it
Added new tests to verify the code changes and built sonic_config_engine-1.0-py3-none-any.whl successfully
Signed-off-by: Arvindsrinivasan Lakshmi Narasimhan <arlakshm@microsoft.com>
In the multi asic platforms all the ASIC are advertising the same IPv6 /64 network from Loopback4096.
Therefore, the IPv6 loopback address of backend asic is not learnt on the frontend asic.
Change the bgpd.conf.main.conf.j2 template file to advertise the Loopback4096 ipv6 address as /128
Why I did it
Quagga is no longer being used. Remove quagga-related code (e.g., docker-fpm-quagga, sonic-quagga, etc.).
How I did it
Remove quagga-related code.
Why I did it
This PR adds changes in sonic-config-engine to consume configuration data in SONiC Yang schema and generate config_db entries
How I did it
Add a new file sonic_yang_cfg_generator .
This file has the functions to
parse yang data json and convert them in config_db json format.
Validate the converted config_db entries to make sure all the dependencies and constraints are met.
Add a new option -Y to the sonic-cfggen command for this purpose
Add unit tests
This capability is support only in sonic-config-engine Python3 package only
Why I did it
portconfig.py gets PORT table from config_db if it is present. If not, port_config.ini files are parsed.
For multi-asic platform, if namespace is passed to get_port_config(), config_db connection was done to host namespace always and not to asic specific namespace.
Provides fix for: #7161
How I did it
Modify db connection function to connect to namespace config_db.
* [TPID CONFIG]TPID attribute Yang model and default TPID for Minigraph to configDB Changes
* Fixed alignment spacing issue, removed length and convert to use tpid_type typedef for both LAG and PORT tpid attribute, added description in port and portchannel leaf for the tpid attribute
Why I did it
The current code skips parsing bandwidth for DeviceMgmtLinks. We have a use case to set the speed for these type of links based on the bandwidth attribute in the minigraph
How to verify it
Ran sonic-cfggen on a minigraph and verified that interface of type DeviceMgmtLink has speed set in the PORT table from the bandwidth attribute in the minigraph
When FECDisabled is set to true in minigraph.py, push 'fec' 'none' explicitly to config_db. When 'fec' is defined in port_config.ini do not override it with 'rs' for 100G
Why I did it
Currently, there is a bug in the ntp.conf jinja2 template where it will ignore the src_intf directive in CONFIG_DB if there are multiple IP addresses associated with an interface. This code change fixes that bug and allows the template to select the correct source interface for NTP.
How I did it
I did this by modifying the macro in ntp.conf.j2 which determines if there is an ip address associated with an interface to set a state variable when it detects a valid interface entry in CONFIG_DB instead of outputting "true" directly (which could result in multiple "trues" outputted for interfaces with multiple valid IP addresses).
How to verify it
Add two ipv4 addresses to an interface in SONiC
Add the following configuration to config_db.json
{
"NTP": {
"global": {
"src_intf": "Ethernet1"
}
}
}
Replace Ethernet1 with the interface name of the one you assigned the IP addresses to.
Run sudo config reload -y
Open /etc/ntp.conf and verify that the following line exists
...
interface listen Ethernet1
...
The interface specified should be the one set in the previous steps.
Description for the changelog
[ntp] Fix ntp.conf template to allow setting of source port in CONFIG_DB
Why I did it
ndppd by default reads /proc/net/ipv6_route ever 30 seconds. Since T1s advertise so many routes to ToRs, this file is extremely large, and reading it causes ndppd's CPU usage to spike every 30 seconds
How I did it
Increase the delay for reading this file to the maximum possible value (max integer value), which will result in CPU spikes every ~24 days instead of every 30 seconds
How to verify it
Start ndppd with the new config file, confirm that no CPU spikes are seen except at startup
Signed-off-by: Lawrence Lee <lawlee@microsoft.com>
This commit contains the following changes to support for configuring a VoQ switch using a minigraph.xml file.:
- Add support for system ports configuration to minigraph
- Add support for SwitchId, SwitchType and MaxCores to minigraph
- Add support for inband vlan configuration in minigraph
- `asic_name` is now a mandatory attribute in CONFIG_DB on VoQ switches
Co-authored-by: Maxime Lorrillere <mlorrillere@arista.com>
This commit has following changes:
* Add templates and code to support VoQ chassis iBGP peers
* Add support to convert a new VoQChassisInternal element in the
BGPSession element of the minigraph to a new BGP_VOQ_CHASSIS_NEIGHBOR
table in CONFIG_DB.
* Add a new set of "voq_chassis" templates to docker-fpm-frr
* Add a new BGP peer manager to bgpcfgd to add neighbors from the
BGP_VOQ_CHASSIS_NEIGHBOR table using the voq_chassis templates.
* Add a test case for minigraph.py, making sure the VoQChassisInternal
element creates a BGP_VOQ_CHASSIS_NEIGHBOR entry, but not if its
value is "false".
* Add a set of test cases for the new voq_chassis templates in
sonic-bgpcfgd tests.
Note that the templates expect the new
"bgp bestpath peer-type multipath-relax" bgpd configuration to be
available.
Signed-off-by: Joanne Mikkelson <jmmikkel@arista.com>
- Add peer_switch field to DEVICE_METADATA table
- In PORT table:
- Set used ports to admin status up
- Set mux_cable to true for downlinks in use
- In MUX_CABLE table:
- Only add entry if the downlink is in use
Signed-off-by: Lawrence Lee <lawlee@microsoft.com>
Make sure Everflow always gets classified as Mirror table and not as Control Plane on multi-asic platforms.
Why I did:
In Multi-asic platforms we generate Everflow acl table data from minigraph for both host and namespace.
It is possible in multi-asic minigraph if there are no external port-channel (Only Router Port IP Interface) then Everflow table will have no binded interface in host and will gets classified as Control Plane ACL while in namespace gets classified as Mirror Table.
For ACL Rule generation we read global db as source of truth for acl table information and so for everflow rule generation if tables gets classified as Control plane we can generate rules with invalid action causing orchagent to throw runtime error.
How I did:
If the table is attach to erspan interface in minigraph then it always gets classified as mirror table.
- When generating L2 preset, check for dual ToR setting from CLI option `-a '{"is_dualtor": true}'`
- When dual ToR is specified, add subtype field to DEVICE_METADATA table
- When dual ToR is specified, add MUX_CABLE, TUNNEL, LOOPBACK_INTERFACE, and PEER_SWITCH tables
Parse the value `AutoNegotiation` element from the `LinkMetadata` section of minigraph file. If the element is present, an `autoneg` key will be added to the port in the `PORT` table of Config DB with a value of either `off` or `on`
If an `autoneg` value is present in port_config.ini, the value from the minigraph will take precedence, overriding that value.
Also remove `AutoNegotiation` and `EnableAutoNegotiation` elements from the `DeviceInfo` section, as we will use this data in the `LinkMetadata` section to determine whether to enable auto-negotiation for a port.
Closes issue #6982.
The issue was root caused as we were using the unix_socket for reading from DB as a default mechanism (#5250). The redis unix socket is created as follows.
admin@str--acs-1:~$ ls -lrt /var/run/redis/redis.sock
srwxrw---- 1 root redis 0 Mar 6 01:57 /var/run/redis/redis.sock
So it used to work fine for the user "root" or if user is part of redis group ( admin was made part of redis group by default )
Check if the user is with sudo permissions then use the redis unix socket, else fallback to tcp socket.
