 |
 |
Hello, We have 4 xrootd servers of differing hardware specs behind a single redirector host. However it appears that, despite some xroot hosts being much more loaded then others, redirections are sent to each node in something like a round-robin manner (each node getting 25% of the transfers, +/- 1%). In my naive understanding I thought that the cms.sched directive oi the redirerector config would be sufficient to enforce some kind of load balancing, and currently this is set to: cms.sched cpu 35 io 30 mem 35 pag 0 runq 0 space 0 fuzz 10 On our xrootd servers, beyond the reference to the manager there is no specific cmsd directives being called, so only: all.manager xgate.hec.lancs.ac.uk:3121 So I was wondering if I was missing something, so perhaps the scheduler was unable to make load-based decisions? The redirector does appears to behave correctly when an xroot server is down for any reason, not redirecting to it. Or perhaps I need to be more aggressive with my cms.sched settings? Any pointers would be appreciated. For reference my redirector and xroot server configs can be seen at: https://github.com/mdoidge/lancsxroot/blob/main/xrootd-redirector.cfg (redirector config) https://github.com/mdoidge/lancsxroot/blob/main/xrootd-cluster.cfg (xrootd server config) (I tried to make an all-in-one config, but it didn't gel for me.) Thanks in advance, Matt ######################################################################## Use REPLY-ALL to reply to list To unsubscribe from the XROOTD-L list, click the following link: https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-L&A=1