Open okt-yurijp opened 10 months ago
@ol-arteman any idea?
It would be helpful to see a bit more of a context, namely, the immediately preceding logs which show the difference between the expected and the actual configuration. However, since /agent/interface/feature/readonly
values are simply taken from ethtool
, my guess is that it's just that a different version of modules is loaded after the reboot.
I performs onload_install before running the test. It installs drivers so after reboot they are loaded automatically, without TE. And, also, sometimes (very occasionally) after reboot, all is fine and further iterations pass.
So do you have any decision?
I tried to use TA reboot TE possibility and faced the following problem, this is error from logs:
ERROR Configurator Self 13:53:15.292 Failed to add/set instance /agent:Agt_A/interface:ens1f1/feature:tx-udp_tnl-segmentation/readonly: (CS-EACCES) RING Configurator Self 13:53:15.293 Sleep 5000 milliseconds to propagate configuration changes 2181 ERROR Configurator Self 13:53:20.293 cfg_backup_verify_and_restore(): failed to restore from the backup: CS-EACCES 2182 ERROR Tester Run 13:53:20.294 Cannot restore configuration backup: CS-EACCES 2183 ERROR Tester Verdict Verdict 13:53:20.294 Internal error
To reproduce this apply sapi-ts-reboot.txt patch to sapi-ts (https://github.com/Xilinx-CNS/cns-sapi-ts), tslib_reboot.txt patch to onload-tslib (https://github.com/Xilinx-CNS/cns-onload-tslib). Also, export the following variable on Test engine: export TE_IUT_REBOOTABLE=yes
export TE_POWER_TA_TYPE=linux_power_ctl_agent export TE_POWER=$TE_ENGINE_HOST export TE_COLD_REBOOT_CMD='path_to_cold_reboot_script %s'
Then, run the following command line: ./run.sh --cfg= --log-html=html --tester-run=sockapi-ts/usecases/read_write%1*100 -q
Then, during the run, you can bring down main interface on Agt_A or do something else to emulate crash on Agt_A. After 2 minutes TE performs cold reboot and then after ~10 minutes TE starts TA once again and then the problem appears.