[dvs] Re-enable RIF tests #1249
Conversation
Signed-off-by: Danny Allen <[email protected]>
|
sonic-net/sonic-sairedis#574 is fixed. |
|
I have been trying to fight this issue for quite a while. it is was first noticed in this pr. sonic-net/sonic-buildimage#4342 judy has spend lots of efforts to track this down. And now I noticed that you have found this issue three weeks ago. |
daall
changed the title
|
I think we need to enable the swss vs tests for sairedis repo. or at least for every sairedis commit, we need to build and build a new swss vs docker and run the test so that we can find the regression. |
I'll grab the latest changes and take a look. Now that you point it out it looks like it's consistently xfailing on |
Agreed. |
|
I think both tests should be fixed. |
|
I fetched the build from sonic-net/sonic-buildimage#4379 and the normal port is consistently passing locally but I'm still hitting the crash with the LAG test. I need to check the email thread with @kcudnik to see if it's the same failure path as before or if this is a different one. |
daall
changed the title
…on (sonic-net#1249) and some other enhancements to techsupport **- What I did** Following is the brief description of the changes, - Adding a ‘--silent’ option to ‘show techsupport’ command. Various tar/untar, addition and removal logs appear on the console by default. This option would disable above logs. - Adding global and per-command timeouts. This would provide more user control on ‘show techsupport’ CLI. - Adding time profiling information for the commands in techsupport. Time profiling information would be part of the tarball and helps to analyse the time consumption per command. - Sometimes ‘syncd’ docker is down and bcmshell is unavailable. In such cases all the bcmcmd commands would timeout and result in tremendous increase in the total techsupport collection time. We provided an option to skip rest of the bcmcmd commands once one command times out. - Added ‘show services’, ‘show reboot-cause’ and various BGP, BFD, bcm shell and other commands - Optimised the /var/log files collection. If the number of files are large in /var/log folder, it takes a long time to add each individually to the tarball. If the folder is tar'ed at once, the time taken reduces significantly. - Following error was observed while tar'ing softlinks inside .etc folder. ** Tar append operation failed. Aborting for safety. ** This issue was due to softlinks present at /etc folder where the destination file is absent. Fixed this issue by deleting such softlinks before adding them to the tarball. **- How I did it** - Added new options to the CLICK command 'show techsupport' - Modified the 'generate_dump' script to accomodate other changes **- How to verify it** Here are some outputs, root@sonic:/home/admin# show techsupport --silent Techsupport is running with silent option. This command might take a long time. HW Mgmt dump script /usr/bin/hw-management-generate-dump.sh does not exist /var/dump/sonic_dump_sonic_20201117_161246.tar.gz root@sonic:/home/admin# root@sonic:~# show techsupport -h Usage: show techsupport [OPTIONS] Gather information for troubleshooting Options: --since TEXT Collect logs and core files since given date -g, --global-timeout INTEGER Global timeout for techsupport in minutes. Default 30 mins -c, --cmd-timeout INTEGER Command timeout for techsupport in minutes. Default 5 mins --verbose Enable verbose output --silent Run techsupport in silent mode -?, -h, --help Show this message and exit. root@sonic:~# **- Previous command output (if the output of a command-line utility has changed)** - Previous command "show techsupport" works as is **- New command output (if the output of a command-line utility has changed)**
Signed-off-by: Danny Allen [email protected]
What I did
Re-enabled most of the RIF tests.
Why I did it
Since we've narrowed down the issue to setting the MTU, we can skip the MTU tests while we implement the fix so that we can at least continue to verify the rest of the RIF behavior.
How I verified it
Local test runs
Details if related