Hi 'Alessandra,
Looking at the "ready" plot , it appears there is no backlog on the fts channel, so issue is to do with how and when atlas add files onto the fts channel. You do have a very high number of concurrent transfers. (25) Dis-regard issues around 0630 to 0730 UTC this morning ( we had internal issue which effected rates.)
IIRC correctly, atlas only put onto a channel files in transferring state when all files associated with all WN jobs associated to task are complete ( hence you will often get a buildup of "transferring files" whilst the system is waiting for the last couple of jobs to finish.)
Brian
-----Original Message-----
From: GRIDPP2: Deployment and support of SRM and local storage management [mailto:[log in to unmask]] On Behalf Of Alessandra Forti
Sent: 16 May 2011 10:07
To: [log in to unmask]
Subject: **Possible spam**fts transfers
Hi,
I changed finally the bonding configuration on Friday. According to my transfers tests things are going much better, but I prefer to let pass few more days to really know, in particular about the sonar tests. In the meantime I'm looking at how jobs are affected and the FTS monitoring attached this morning ganglia plot. Over the weekend we have accumulated a big number of jobs in transfering state and I know there was a general problem with dq2 on Saturday but I'd have expected a continuous flow to clear the data instead there are a number of high peaks. Is there is a reason for these regular ups and downs rather than continuous transfers?
cheers
alessandra
|