 |
 |
Thanks Andy, Coming in after the weekend we can see load "reports" arriving every ~2 hours and balancing decisions are being made. It appears I just needed to be patient! We've still got tuning to be done on this, but we're no longer stalled by missing configrations. Thanks again, Matt ________________________________________ From: Andrew Hanushevsky <[log in to unmask]> Sent: 20 January 2023 23:41 To: Doidge, Matt Cc: [log in to unmask] Subject: [External] Re: cmsd.sched not (noticeably) load balancing redirection This email originated outside the University. Check before clicking links or attachments. Hi Matt, The warning is issued, I suspect, is because the cms.sched directive was restricted to the manager (i.e. redirector). It should not be as it applies to the redirector and data servers. The recoding of the perf information occurs periodically and not atthe interval that it is reported. This is to keep the log file at a reasonable size. See the cms.ping directive for more information.... https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fxrootd.slac.stanford.edu%2Fdoc%2Fdev54%2Fcms_config.htm%23_Toc53611094&data=05%7C01%7Cdoidgem%40live.lancs.ac.uk%7C6fbcf6fef7be4cb29e8408dafb3fd65e%7C9c9bcd11977a4e9ca9a0bc734090164a%7C0%7C0%7C638098548849069401%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=TD5B%2B49wiu69dNETq6huSblSV9mDZjd5YyhL6WE4SbY%3D&reserved=0 As for only one node reporting information, make sure each data server has been properly configured and restarted. Andy On Fri, 20 Jan 2023, Doidge, Matt wrote: > Thank you Andy and Max for your replies, which have put me on the right track. > > I've tried to implent the perf directive using the supplied cms_monPerf utility by sticking at the end of our config: > if exec cmsd > # call preinstall script every minute > cms.perf int 1m pgm /usr/share/xrootd/utils/cms_monPerf 60 > fi > > Within the logs however I see a config warning I do not understand at all: > " > =====> cms.perf int 1m pgm /usr/share/xrootd/utils/cms_monPerf 60 > The following paths are available to the redirector: > w /cephfs/grid > > Config warning: metrics supplier specified without any scheduling metrics! > ------ [log in to unmask] phase 1 server initialization completed. > " > > And it does not appear that metrics are being fed to our manager host (although cms_monPerf is running on our servers). Looking in the logs we only have one instance of metric information arriving in our manager's cmsd.log, lines like: > 230120 14:04:26 259335 Node: stor015.hec.lancs.ac.uk load=30; cpu=6 net=99 inq=45 mem=3 pag=0 dsk=2147483647 utl=74 shr=[100 26 0] ref=[10906 1721] > > These metrics were delivered once, about 10 minutes after the servers were restarted and we've not seen any mentions in the logs since. I'm not sure if this is a data point or just noise though. Data is still flowing to the xrootd servers though, in the round-robin fashion. > > My apologies if I'm missing something as equally straightforward as with my original mail, but I'd appreciate any pointers. > > Thanks again, and have a good weekend all, > Matt > > ________________________________________ > From: Fischer, Max (SCC) <[log in to unmask]> > Sent: 20 January 2023 08:06 > To: Doidge, Matt > Cc: [log in to unmask] > Subject: [External] Re: cmsd.sched not (noticeably) load balancing redirection > > This email originated outside the University. Check before clicking links or attachments. > > ######################################################################## > Use REPLY-ALL to reply to list > > To unsubscribe from the XROOTD-L list, click the following link: > https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Flistserv.slac.stanford.edu%2Fcgi-bin%2Fwa%3FSUBED1%3DXROOTD-L%26A%3D1&data=05%7C01%7Cdoidgem%40live.lancs.ac.uk%7C6fbcf6fef7be4cb29e8408dafb3fd65e%7C9c9bcd11977a4e9ca9a0bc734090164a%7C0%7C0%7C638098548849069401%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=lnGr7A7TWYEL893alRZaGpW%2Bx9szkjpCmEtZsnCHdG8%3D&reserved=0 > ######################################################################## 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