* [swss] Chassis db clean up optimization and bug fixes
This commit includes the following changes:
- Fix for regression failure due to error in finding CHASSIS_APP_DB in
pizzabox (#PR 16451)
- After attempting to delete the system neighbor entries from
chassis db, before starting clearing the system interface entries,
wait for sometime only if some system neighbors were deleted.
If there are no system neighbors entries deleted for the asic coming up,
no need to wait.
- Similar changes for system lag delete. Before deleting the
system lag, wait for some time only if some system lag memebers were
deleted. If there are no system lag members deleted no need to wait.
- Flush the SYSTEM_NEIGH_TABLE from the local STATE_DB. While asic
is coming up, when system neigh entries are deleted from chassis ap
db (as part of chassis db clean up), there is no orchs/process running to
process the delete messages from chassis redis. Because of this, stale system
neigh are entries present in the local STATE_DB. The stale entries result in
creation of orphan (no corresponding data path/asic db entry) kernel neigh
entries during STATE_DB:SYSTEM_NEIGH_TABLE entries processing by nbrmgr (after
the swss serive came up). This is avoided by flushing the SYSTEM_NEIGH_TABLE from
the local STATE_DB when sevice comes up.
Signed-off-by: vedganes <veda.ganesan@nokia.com>
* [swss] Chassis db clean up bug fixes review comment fix - 1
Debug logs added for deletion of other tables (SYSTEM_INTERFACE and SYSTEM_LAG_TABLE)
Signed-off-by: vedganes <veda.ganesan@nokia.com>
---------
Signed-off-by: vedganes <veda.ganesan@nokia.com>
(cherry picked from commit b13b41fc22)
chassis-packet: Update arp_update script for FAILED and STALE check (#16311)
1. Fixing an issue with FAILED entry resolution retry.
Neighbor entries in arp table may sometimes enter a FAILED state when the far end is down and reports the state as follows:
2603:10e2:400:3::1 dev PortChannel19 router FAILED
While the arp_update script handles the entries for FAILED in the following format, the above was not handled due to the token location (extra router keyword at index 4):
2603:10e2:400:3::1 dev PortChannel19 FAILED
The former format may appear if an arp resolution is tried on a link that is known but the far end goes down, e.g., pinging a STALE entry while the far end is down.
2. Refreshing STALE entries to make sure the far end is reachable.
STALE entries for some backend ports may appear in chassis-packet when no traffic is received for a while on the port. When the far end goes down, it is expected for BFD to stop sending packets on the session for which the far end is not reachable. But as the entry is known as stale, on the Cisco chassis, BFD keeps sending packets. Refreshing the stale entry will keep active links as reachable in the neighbor table while the entries for the far end down will enter a failed state. FAILED state entries will be retired and entered reachable when far end comes back up.
* [chassis]Chassis DB cleanup when asic comes up
Cleanup the entries from the following tables in chassis app db in
redis_chassis server in the supervisor
(1) SYSTEM_NEIGH
(2) SYSTEM_INTERFACE
(3) SYSTEM_LAG_MEMBER_TABLE
(4) SYSTEM_LAG_TABLE
As part of the clean up only those entries created by the asic that
is coming up are deleted. The LAG IDs used by the asics are also
de-allocated from SYSTEM_LAG_ID_TABLE and SYSTEM_LAG_ID_SET
- Added check to run the chassis db clean up only for voq switches.
Signed-off-by: vedganes <veda.ganesan@nokia.com>
Co-authored-by: vganesan-nokia <67648637+vganesan-nokia@users.noreply.github.com>
Fixes#15667 and #13293
Work item tracking
Microsoft ADO 24472854:
How I did it
On chassis supervisor bgp feature is disabled in hostcfgd. The dependency between swss and bgp causes the bgp containers to start even though the feature is disabled.
How to verify it
Tests on chassis supervisor and LC
Co-authored-by: Arvindsrinivasan Lakshmi Narasimhan <55814491+arlakshm@users.noreply.github.com>
* Fix CONFIG_DB_INITIALIZED flag check logic and set/reset flag for warm-reboot
* Fix db-cli usage
* Handle same image warm-reboot and generalize handling of INIT flag
* Cover boot from ONIE case: set config init flag when minigraph, config_db are missing
* Handle case: first boot of SONiC
* Check for config init flag
* Simplify logic, and do not call db_migrator for same image reboot
Co-authored-by: Vaibhav Hemant Dixit <vaibhav.dixit@microsoft.com>
Why I did it
The hw resources should be released before updating firmware.
How I did it
Added logic to release hw resources in syncd.sh script
Signed-off-by: Vadym Hlushko <vadymh@nvidia.com>
Co-authored-by: Vadym Hlushko <62022266+vadymhlushko-mlnx@users.noreply.github.com>
Why I did it
Fixes#14179
chassis-packet: missing arp entries for static routes causing high orchagent cpu usage
It is observed that some sonic-mgmt test case calls sonic-clear arp, which clears the static arp entries as well. Orchagent or arp_update process does not try to resolve the missing arp entries after clear.
How I did it
arp_update should resolve the missing arp/ndp static route
entries. Added code to check for missing entries and try ping if any
found to resolve it.
How to verify it
After boot or config reload, check ipv4 and ipv4 neigh entries to make sure all static route entries are present
manual validation:
Use sonic-clear arp and sonic-clear ndp to clear all neighbor entries
run arp_update
Check for neigh entries. All entries should be present.
Testing on T0 setup route/for test_static_route.py
The test set the STATIC_ROUTE entry in conifg db without ifname:
sonic-db-cli CONFIG_DB hmset 'STATIC_ROUTE|2.2.2.0/24' nexthop 192.168.0.18,192.168.0.25,192.168.0.23
"STATIC_ROUTE": {
"2.2.2.0/24": {
"nexthop": "192.168.0.18,192.168.0.25,192.168.0.23"
}
},
Validate that the arp_update gets the proper ARP_UPDATE_VARDS using arp_update_vars.j2 template from config db and does not crash:
{ "switch_type": "", "interface": "", "pc_interface" : "PortChannel101 PortChannel102 PortChannel103 PortChannel104 ", "vlan_sub_interface": "", "vlan" : "Vlan1000", "static_route_nexthops": "192.168.0.18 192.168.0.25 192.168.0.23 ", "static_route_ifnames": "" }
validate route/test_static_route.py testcase pass.
arp_update should resolve the missing arp/ndp static route
entries. Added code to check for missing entries and try ping to
resolve the missing entry.
Why I did it
Fixes#14179
chassis-packet: missing arp entries for static routes causing high orchagent cpu usage
It is observed that some sonic-mgmt test case calls sonic-clear arp, which clears the static arp entries as well. Orchagent or arp_update process does not try to resolve the missing arp entries after clear.
How I did it
arp_update should resolve the missing arp/ndp static route
entries. Added code to check for missing entries and try ping if any
found to resolve it.
How to verify it
After boot or config reload, check ipv4 and ipv4 neigh entries to make sure all static route entries are present
manual validation:
Use sonic-clear arp and sonic-clear ndp to clear all neighbor entries
run arp_update
Check for neigh entries. All entries should be present.
Signed-off-by: anamehra <anamehra@cisco.com>
Why I did it
Vendor platform may require running platform specific pre-shutdown routine before shutting down the syncd process which runs the SAI and vendor sdk instance.
How I did it
Added a platform script hook which will be executed if the plugin script is provided by the platform in device//plugins/
Changing the default config knob value to be True for killing radv, due to the reasons below:
Killing RADV is to prevent sending the "cease to be advertising interface" protocol packet.
RFC 4861 says this ceasing packet as "should" instead of "must", considering that it's fatal to not do this.
In active-active scenario, host side might have difficulty distinguish if the "cease to be advertising interface" is for the last interface leaving.
6.2.5. Ceasing To Be an Advertising Interface
shutting down the system.
In such cases, the router SHOULD transmit one or more (but not more
than MAX_FINAL_RTR_ADVERTISEMENTS) final multicast Router
Advertisements on the interface with a Router Lifetime field of zero.
In the case of a router becoming a host, the system SHOULD also
depart from the all-routers IP multicast group on all interfaces on
which the router supports IP multicast (whether or not they had been
advertising interfaces). In addition, the host MUST ensure that
subsequent Neighbor Advertisement messages sent from the interface
have the Router flag set to zero.
sign-off: Jing Zhang zhangjing@microsoft.com
Co-authored-by: Jing Zhang <zhangjing@microsoft.com>
Why I did it
Limiting #12804 changes to PikeZ platform only (Arista-720DT-48S). Note that this is a short term workaround for this platform until SDK investigation on SDK init failure on docker syncd restart due to DMA issues is resolved.
How I did it
Retrieve platform name from /host/machine.conf and only reload SDK kmods on Arista-720DT-48S platform.
Signed-off-by: Michael Li <michael.li@broadcom.com>
- Why I did it
There's a slowdown in bootup related to the execution of a show command during startup of swss service. show is a pretty heavy command and takes long time to execute ~2 sec.
- How I did it
I replaced show with sonic-db-cli which takes a ms to run.
- How to verify it
Boot the switch and verify swss is active.
Signed-off-by: Stepan Blyschak <stepanb@nvidia.com>
Why I did it
There is an issue on the Arista PikeZ platform (using T3.X2: BCM56274) while running SONiC. If the 'syncd' container in SONiC is restarted, the expected behaviour is that syncd will automatically restart/recover; however it does not and always fails at create_switch due to BCM SDK kmod DMA operation cancellation getting stuck.
Sep 16 22:19:44.855125 pkz208 ERR syncd#syncd: [none] SAI_API_SWITCH:platform_process_command:428 Platform command "init soc" failed, rc = -1. Sep 16 22:19:44.855206 pkz208 INFO syncd#supervisord: syncd CMIC_CMC0_PKTDMA_CH4_DESC_COUNT_REQ:0x33#015 Sep 16 22:19:44.855264 pkz208 CRIT syncd#syncd: [none] SAI_API_SWITCH:platformInit:1909 initialization command "init soc" failed, rc = -1 (Internal error). Sep 16 22:19:44.855403 pkz208 CRIT syncd#syncd: [none] SAI_API_SWITCH:sai_driver_init:642 Error initializing driver, rc = -1. ... Sep 16 22:19:44.855891 pkz208 CRIT syncd#syncd: [none] SAI_API_SWITCH:brcm_sai_create_switch:1173 initializing SDK failed with error Operation failed (0xfffffff5).
Reloading the BCM SDK kmods allows the switch init to continue properly.
How I did it
If BCM SDK kmods are loaded, unload and load them again on syncd docker start script.
How to verify it
Steps to reproduce:
In SONiC, run 'docker ps' to see current running containers; 'syncd' should be present.
Run 'docker stop syncd'
Wait ~1 minute.
Run 'docker ps' to see that syncd is missing.
Check logs to see messages similar to the above.
Signed-off-by: Michael Li <michael.li@broadcom.com>
Fix the issue where arp_update will not ping some of the ip's even
though they are in failed state since grep of that ip on ip neigh show
command does not do exact word match and can return multiple match.
Why I did it
If the SWSS services was restarted, the MACsec service should also be restarted. Otherwise the data in wpa_supplicant and orchagent will not be consistent.
How I did it
Add dependency in docker-macsec.mk.
How to verify it
Manually check by 'sudo service swss restart'.
The MACsec container should be started after swss, the syslog will look like
Sep 8 14:36:29.562953 sonic INFO swss.sh[9661]: Starting existing swss container with HWSKU Force10-S6000
Sep 8 14:36:30.024399 sonic DEBUG container: container_start: BEGIN
...
Sep 8 14:36:33.391706 sonic INFO systemd[1]: Starting macsec container...
Sep 8 14:36:33.392925 sonic INFO systemd[1]: Starting Management Framework container...
Signed-off-by: Ze Gan <ganze718@gmail.com>
* [mux] skip mux operations during warm shutdown
- Enhance write_standby.py script to skip actions during warm shutdown.
- Expand the support to BGP service.
- MuX support was added by a previous PR.
- don't skip action during warm recovery
Signed-off-by: Ying Xie <ying.xie@microsoft.com>
After pinging any failed IPv6 neighbor entries, set the remaining failed/incomplete entries to a permanent INCOMPLETE state. This manual setting to INCOMPLETE prevents these entries from automatically transitioning to FAILED state, and since they are now incomplete any subsequent NA messages for these neighbors is able to resolve the entry in the cache.
Signed-off-by: Lawrence Lee <lawlee@microsoft.com>
As part of PR #11754
Change was added to use variable SONIC_DB_NS_CLI for
namespace but that will not work since ./files/scripts/syncd_common.sh
uses SONIC_DB_CLI. So revert back to use SONIC_DB_CLI and define new
variable for SONIC_GLOBAL_DB_CLI for global/host db cli access
Also fixed DB_CLI not working for namespace.
Change `sxdkernel start` to `sxdkernel restart`. If `syncd` service crashes in `ExecStartPre` systemd will not call `ExecStop` and thus will not call `sxdkernel stop`. Use of `sxdkernel restart` is more robust in terms of guarantees to restore the system after unexpected crashes.
Signed-off-by: Stepan Blyschak <stepanb@nvidia.com>
Signed-off-by: Stepan Blyschak <stepanb@nvidia.com>
Why I did:
In case of multi-asic platforms gbsyncd is not getting added to Feature Table of Host Config DB. Without this container_checker complains of not needed gbsyncd container's are running.
How I did:
Update Both Host and Namespace config db when gbsyncd docker is starting.
How I verify:
Verified on Multi-asic platforms.
bgp should be a per-asic service, and runs for each namespace on
multi-asic platforms. However, putting bgp in MULTI_INST_DEPENDENT
causes swss to be restarted as well as bgp. this is causing issues after #11000
Issue: #11653
This fix:
removes bgp from dependents list
adds a conditional that either adds bgp, or bgp@$DEV to separate
between single and multi-asic platforms
In arp_update, check for FAILED or INCOMPLETE kernel neighbor entries and manually ping them to try and resolve the neighbor
Signed-off-by: Lawrence Lee <lawlee@microsoft.com>
When using trap on SIGTERM the script will not react to the SIGTERM signal sent while a child is executing.
I.e, the following script does not react on SIGTERM sent to it if it is
waiting for sleep to finish:
```
trap "echo Handled SIGTERM" 0 2 3 15
echo "Before sleep"
sleep inf
echo "After sleep"
```
Instead, trap only on EXIT which covers also a scenario with exit on
SIGINT, SIGTERM.
Signed-off-by: Stepan Blyschak <stepanb@nvidia.com>
Why I did it
The initial value has to be present for the state machines to work. In active-standby dual-tor scenario, or any hardware mux scenario, the value will be updtaed eventually with a delay.
However, in active-active dual-tor scenario, there is no other mechanism to initialize the value and get state machines started.
So this script will have to write something at start up time.
For active-active dualtor, 'active' is a more preferred initial value, the state machine will switch the state to standby soon if
link prober found link not in good state.
How I did it
Update the script to always provide initial values.
How to verify it
Tested on active-active dual-tor testbed.
Signed-off-by: Ying Xie ying.xie@microsoft.com
A change in sonic-utilities makes all cache files be saved into a
/tmp/cache. On swss restart this cache has to be removed in case swss
starts in cold or fast mode. A related cache restoration in the warmboot
finalizer script is also updated to use new location.
- Why I did it
To fix#9817. Clear the cache directory on swss.sh except for warm start.
Also, adopted finalize-warmboot script to take the cache directory.
- How I did it
A change in sonic-utilities makes all cache files be saved into a /tmp/cache. On swss restart this cache has to be removed in case swss starts in cold or fast mode. A related cache restoration in the warmboot finalizer script is also updated to use new location.
- How to verify it
Run togather with Azure/sonic-utilities#2232. Verify counters cache is removed on config reload, cold/fast reboots, swss restart.
Signed-off-by: Stepan Blyschak <stepanb@nvidia.com>
What I did:
Added bgp as a dependent of swss
Why I did it:
bgp container was not restarting on swss crash. When swss crashes, linkmgrd
doesn't initate a switchover because it cannot access the default route from
orchagent. Bringing down bgp with swss will isolate the ToR, causing linkmgrd
to initiate a switchover to the peer ToR avoiding significant packet loss.
How I did it:
Added bgp to DEPENDENT
Signed-off-by: Nikola Dancejic <ndancejic@microsoft.com>
What I did:
Following changes done for packet based chassis:-
1> Run arp_update on LC's to resolve static route nexthops over backend
port-channel interfaces.
2> On Supervisor make sure arp_update exit gracefully
Avoid write_standby in warm restart context.
sign-off: Jing Zhang zhangjing@microsoft.com
Why I did it
In warm restart context, we should avoid mux state change.
How I did it
Check warm restart flag before applying changes to app db.
How to verify it
Ran write_standby in table missing, key missing, field missing scenarios.
Did a warm restart, app db changes were skipped. Saw this in syslog:
WARNING write_standby: Taking no action due to ongoing warmrestart.
- Why I did it
When LLDP is disabled through feature command, it gets spawned after reboot.
- How I did it
In syncd.sh check if the service is enabled before spawning automatically during cold reboot.
- How to verify it
Disable lldp feature. Perform cold reboot and verify its not spawned.
- Why I did it
Recent change to delay PMON service in case of fast/warm reboot introduce an issue when restarting only SWSS service after fast/warm reboot for Nvidia platform.
Since the timer is triggered only when the system boot, in a scenario when the system is after a fast/warm reboot and the user restart SWSS service, as part of syncd.sh script, PMON service will stop but the timer will not start again.
- How I did it
On syncd.sh script, in case of fast/warm indication, check if pmon.timer is running.
If it is running it means we are at the first boot and continue normally.
If it is not running, meaning the service was restarted, start the timer to keep the system behavior consistent.
- How to verify it
Run fast/warm reboot.
service swss restart.
Observe PMON service starting.
Signed-off-by: Shlomi Bitton <shlomibi@nvidia.com>
* When reloading config after crashes, VTEP interfaces are sometimes not created since the tunnel still exists in the STATE_DB.
* Adding VXLAN_TUNNEL_TABLE to the list of tables to be cleaned in swss.sh fixes the problem.
- Why I did it
Profiling the system state on init after fast-reboot during create_switch function execution, it is possible to see few python scripts running at the same time.
This parallel execution consume CPU time and the duration of create_switch is longer than it should be.
Following this finding, and the motivation to ensure these services will not interfere in the future, PMON is delayed in 90 seconds until the system finish the init flow after fastboot.
- How I did it
Add a timer for PMON service.
Exclude for MLNX platform the start trigger of PMON when SYNCD starts in case of fastboot.
Copy the timer file to the host bin image.
- How to verify it
Run fast-reboot on MLNX platform and observe faster create_switch execution time.
- Why I did it
Profiling the system state on init after fast-reboot during create_switch function execution, it is possible to see few python scripts running at the same time.
This parallel execution consume CPU time and the duration of create_switch is longer than it should be.
Following this finding, and the motivation to ensure these services will not interfere in the future, LLDP is delayed in 90 seconds until the system finish the init flow after fastboot.
- How I did it
Add a timer for LLDP service.
Copy the timer file to the host bin image.
- How to verify it
Run fast-reboot on MLNX platform and observe faster create_switch execution time.
This PR is dependent on PR: #10567
