An introduction was given by Sam Skispey:
DOME is the brain child of Fabrizio Furano during his turn in charge of DPM. DPM Code base is frequently considered hard to maintain because it is written by a single developer who then hands it over to someone else. DOME is a fairly radical change under the hood:
- It changes a lot of the process running with an XRootD system with plugins.
- It does not change the database.
- DOME doesn’t understand the SRM.
- It is possible to run DOME + legacy services to enable the SRM to still run, however we want to move to pure DOME sites.
Clarifications from Matt's talk:
After September, there will be no support (including security) for legacy DPM. You will need to be running pure DOME. The timing of site upgrades is being coordinated in the storage meeting and won't be repeated here.
Enabling gridftp_redirect => true is critically important. Otherwise all gridftp transfers go through your headnode!
Tim gave very clear instructions on how to get ATLAS to stop using the SRM.
Actions:
Sam Skipsey to create tools that work in DOME that will assist a site admin in the event that a disk server is lost.
Alastair Dewhurst to send out an email and ask sites which VOs they support. We can then compile a list of instructions on how we get each VO to stop using the SRM. (Wiki page created at: https://www.gridpp.ac.uk/wiki/Retiring_SRM)
There are minutes attached to this event.
Show them.