dCache Tier I meeting August 9, 2011

[part of a series of meetings]

Present

dCache.org(Antje,Tanja,Patrick), IN2P3(Nicolas), Sara(), Triumf(), BNL(), NDGF(), PIC(), GridKa(Doris), Fermi(), CERN()

Agenda

Remaining question from Doris from last week.

Q: What does 'rc retry same host' mean.

A: Options are 'never', 'besteffort', notchecked. This is for file restore from tape. If the first restore fails, the pool manager tries again. The host it choses for the retry depends on this setting. 'never' it will never try the same host again. 'best effort' it will try another host but only if one is available. 'not checked' it will do the same cost calculation as before and assigns the result of the cost calculation independently of the selection before.

Question 1 by Nicolas:

Q: Do we need loadbalancing on the dcap door.

A: Is not only beneficial for load balancing but as well to avoid single point of failures. So at least a second door would be good. IN2P3 is using a Proxy system. But could as well use DNS or F5 rebalancing.

Question 2 by Nicolas:

Q: Some dcap door remain for days. The only way for Nicolas is to restart the door.

A: Please check the netstat status of the connection and check if the client is still active only the workernode and open a ticket if not.

Site reports

PIC : Upgraded to 1.9.12 and observed a problem with FTP write triggered by SRM to SRM. Tigran and Gerd investigated and Tigran will provide a patch within the next 1/2 hour.

Support tickets for discussion

[Items are added here automagically]

DTNM

Proposed: same time, next week.