Closed JBlack93 closed 4 years ago
Is there a reason to implement this? @jcwhitehead already added a (incremental) sleep time in #43, together with other changed to the copy. The only thing missing is removing root
and xrootd
from the protocol list.
I would argue that this is a quick and dirty fix to the problem presented in #44.
Not ideal, and to be superseded by #43.
Given the nature of the issue I think a quick patch is worth implementing while the new copy is validated.
Redundant after #43
Firstly: Removal of protocols that do not work. (Suggestion from Adam and Paul):
Secondly: Dramatically increase wait time upon failed copy to ensure that the repeated copy attempts do not hamper the server with an increased load (also randomise this wait time, to ensure that we avoid distributing a synchronised set of jobs which failed on the same copy command) See issue #44 .
Waiting ~half a minute for a job that takes 4+ hours shouldn't be a problem.