Added by pgeorgiadis over 6 years ago. Updated over 6 years ago.
100%
Description
Thorsten needs at minimum '/var/log/transactional-update.log' or the output of transactional update with the error.
I am thinking to collect all of them in the stack_controller, after the update, in the bash script to add:
docker exec -i $(docker ps | grep salt-master | awk '{print $1}') salt --batch 15 -P "roles:(admin)" cmd.run 'cat /var/log/transactional-update.log' > transactional_admin.log
docker exec -i $(docker ps | grep salt-master | awk '{print $1}') salt --batch 15 -P "roles:(kube-master)" cmd.run 'cat /var/log/transactional-update.log' > transactional_masters.log
docker exec -i $(docker ps | grep salt-master | awk '{print $1}') salt --batch 15 -P "roles:(kube-minion)" cmd.run 'cat /var/log/transactional-update.log' > transactional_minions.log
tar zcvf transactional-update.tar.gz transactional_*.log
what do you think?
You need to upload_logs after that anyway, so I think we can do it directly (not from controller)
https://github.com/os-autoinst/os-autoinst-distri-opensuse/pull/4014
if something breaks in Velum in the meantime, it will never manage to get to the mutex_create 'UPDATE_FINISHED'; and the handle_update will never reach export_logs. As a result the the logs will never be uploaded in case of Velum failure. I would like to avoid that and having the logs update the actual transactual update (before switching to velum). This can be done with mutexes if you want per/node logs, or aggregate them and collect them at the controller right before switching to Velum. What do you think?
thx