 |
 |
On 01/31/2014 01:49 AM, Andrew Hanushevsky wrote: > Hi Adrian, Hi! > The redirector really has no good way of adding up all the statistics. > In fact, any attempt to do such a thing is rather non-scalable since > clusters are of indefinite size. In cases like these, we defer to the make sense .. now that you mention this i imagine that would be a burden on redirector to keep asking for sizes and add them (for up to 64 data servers maybe .. but for more clearly would bring latency in data serving) ... > client (much like listing a directory) where each server if asked for > its space statistics and the client adds all of them up before > presenting the information. That would be relatively easy to do except > that space statistics are implementation dependent and it would be > pretty difficult for the client to handle all possible formats. I > suppose it could handle only the default "oss-style" format as that is > documented (http://xrootd.org/doc/prod/xrd_monitoring.htm#_Toc346032364) this would be great! Thank you!! Adrian > and let the users force other storage providors to use the same format. > Lukasz? > > Andy > > -----Original Message----- From: Lukasz Janyst > Sent: Thursday, January 30, 2014 3:25 AM > To: Adrian Sevcenco ; [log in to unmask] > Subject: Re: xrdfs usage (howto find used/free/total space) > > On 30.01.2014 12:18, Adrian Sevcenco wrote: >>> The output of 'query space' is implementation dependent. In pure >>> xrootd you get: >>> >>> ]==> xrdfs metaman query space /data 2 13 31 2 13 31 >>> >>> In EOS you get: >>> >>> ]==> xrdfs eosatlas query space / >>> oss.cgroup=default&oss.space=18747277867692032&oss.free=4341978005811200&oss.maxf=68719476736&oss.used=14405299861880832&oss.quota=18747277867692032 >>> >> hmm, this would be nice to get .. > > Andy? > >>> Isn't it what you need, except that in MB and %? >> nope .. i have 222 TB of data (now is less because of down dataserver).. >> you can see details here: >> http://alimonitor.cern.ch/stats?filter_0=ISS&page=SE%2Ftable >> >> So, the output is not correct.. > > It is correct, it just returns the info about the manager node and > not the whole cluster. > >> the thing is that i understood that xrdfs should be the preferred method >> for xrood interrogation so this is why i try with xrdfs.. > > You're correct. > >> in the mean time i will use this: >> echo exit | ~/xrdserver/bin/xrd $1.spacescience.ro queryspace / - >> >> as it give me some information > > Again, thanks for pointing this out. This command does nothing more > than recursively sending the 'query space' to all the nodes in your > cluster and summing up the results. It is however unreliable, because it > does not take into account the implementation differences. We will look > into implementing this in a reliable way. > > Cheers, > Lukasz > > ######################################################################## > 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 ######################################################################## 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