Save pods' log before deleting that

[kebe7jun]

Now, when a Pod of a job fails, the entire task will be restarted (Pod deleted), which will cause log loss and cannot be traced back. We need a solution to save these logs.