Choose timezone
Your profile timezone:
ADC Weekly
http://indico.cern.ch/conferenceDisplay.py?confId=259107
MichaelErnst SIR
AndrejFilipic:
discussion
-one of the outlook point is about having more copies of "fresh" data (which are the most problematic). Has been pointed out that this could help if they are at the same site, multiple copy at different sites it's early to say it, because the recovery service didn't finish yet its job, so even now we don't know how many files have multiple copies (but still some urgent tasks needed to be finished as quickly as possible)
- comment on recovery services (Cedric) : the present recovery service has been designed to handle 30k/60k file losses, this is one order of magnitude more. Improvements in the present service are not foreseen, but the Rucio recovery service will be thought to handle also those kind of disasters. - other comment from Ale: next time we have to remember to wait before injecting all the bunch of files, we may decide before which are more important.
DarioBarberis:
Q:can we do a copy and then a delete (few days/week after) instead of just a move?
Dario: will check, maybe it's possible.
Andreu:
- agreed to set 99.9% 15h
- agreed no memory limits, physical and virtual can create issues.
--> to be agreed if and which announcements we will do.
Q(Alastair) can we define on the site 2 analysis queue, one long, one short.
Answer: yes, behind 2 different PandaResources
AOB
Dario: I see a DB entry - Ale: this was just a clone from the previous meeting, sorry for this - Dario: I take the opportunity to tell that there is the request from IT-DB to close outside connections to ATLAS DB. will circulate email with more details.
MichaelErnst SIR
AndrejFilipic:
discussion
-one of the outlook point is about having more copies of "fresh" data (which are the most problematic). Has been pointed out that this could help if they are at the same site, multiple copy at different sites it's early to say it, because the recovery service didn't finish yet its job, so even now we don't know how many files have multiple copies (but still some urgent tasks needed to be finished as quickly as possible)
- comment on recovery services (Cedric) : the present recovery service has been designed to handle 30k/60k file losses, this is one order of magnitude more. Improvements in the present service are not foreseen, but the Rucio recovery service will be thought to handle also those kind of disasters. - other comment from Ale: next time we have to remember to wait before injecting all the bunch of files, we may decide before which are more important.
DarioBarberis:
Q:can we do a copy and then a delete (few days/week after) instead of just a move?
Dario: will check, maybe it's possible.
Andreu:
- agreed to set 99.9% 15h
- agreed no memory limits, physical and virtual can create issues.
--> to be agreed if and which announcements we will do.
Q(Alastair) can we define on the site 2 analysis queue, one long, one short.
Answer: yes, behind 2 different PandaResources
AOB
Dario: I see a DB entry - Ale: this was just a clone from the previous meeting, sorry for this - Dario: I take the opportunity to tell that there is the request from IT-DB to close outside connections to ATLAS DB. will circulate email with more details.
There are minutes attached to this event.
Show them.