[Integration] Modify dirac-admin-get-pilot-output to get remote pilot log

[martynia]

BEGINRELEASENOTES

*WMS
NEW: Enhance dirac-admin-get-pilot-output to download remote pilot logs. Currently only from a SE.

ENDRELEASENOTES

Changes are:

1. dirac_admin_get_pilot_output.py - add a -r switch to get remote pilot logs rather than classic ones.
2. DiracAdmin.py to call a service method for remote logs
3. PilotManagerHandler.py to load a plugin for pilot logs retrieval.
4. FileCacheDownloadPlugin.py a plugin to download a log file and return it to a client

[fstagni]

Somehow it does not feel right that user would have to decide about this (seems too low level to be exposed at API level)

[martynia]

What would be an other option? If it is used in an Dirac-admin command then it is an expert route. The intention was to have a classic log retrieval mechanism a default one, and the external log retriever only used in special cases, no?

[fstagni]

Then things can be tried in this order?

[martynia]

We could try to get logs in a specific order: classic then remote. The specific order could even be configurable on the server. The point is, that if someone ever wants to test the remote log retrieval and the classic logs come first, then there is no way to get the remote logs by using a dirac-admin command line if we drop the -r flag.

So my proposal is:

• remove the -r flag
• set the default order to classic/remote
• add a config flag to the CS (remoteLogsPriority=True) or similar. Default False.

[fstagni]

You would need to rebase and fix the conflicts.

[martynia]

I rebased, but there are some strange errors I don't recognise...

[fstagni]

Is this really an LFN? Looks like just a path to a file to me.

[martynia]

It is. It is passed to getFile later

[fstagni]

I am afraid we have been using this pattern in other places, but os.path is not the most precise way to manipulate/create LFNs, as this is OS-dependent. We might argue that we only run on Linux so it does not really matter, but more precise would be to at least to use posixpath instead (@chaen and @atsareg might comment)