 |
 |
Hi Andy again, I have found that stupid thing. And it is magic end of line :)) Why it was not taken the particular line with supervisor is because the string "supervisor" in if statement had additional blank characters and I didn't see that. Maybe it would be good to demark the statement in brackets to beware of it. Sorry to spam you and thanks for help Pavel Pavel Jakl wrote: > Hi Andy, > > I re-produced it again. As you could see from the olb log file: > > 060328 17:40:22 001 (c) 2005 SLAC olbd version 20060307-0546 > > 060328 17:40:22 001 [log in to unmask] initilization > started. > > 060328 17:40:22 001 olb_Config: Role not specified; manager role assumed. > > 060328 17:40:22 19119 olb_Xeq: Performance monitor thread started > > 060328 17:40:22 19119 olb_Xeq: Refcount monitor thread started > > 060328 17:40:22 19119 olb_Xeq: Request Responder thread started > > 060328 17:40:22 19119 olb_Xeq: Request Timeout thread started > > 060328 17:40:22 19119 olb_Xeq: State monitor thread started > > 060328 17:40:22 001 [log in to unmask]:3121 manager > initialization completed. > > 060328 17:40:22 19119 olb_Xeq: Admin traffic thread started > > 060328 17:40:27 19119 olb_Admin_Login: Primary server 19060 logged in > > 060328 17:41:52 19119 olb_Xeq: Worker thread started > > 060328 17:41:52 19119 olb_Config: Service enabled. > > And please also see that this node has as instance name "supervisor" > and I have these lines in configuration file: > > olb.role manager if named redirector > > olb.role server if named dataserver > > olb.role supervisor if named supervisor > > > I am also sending the whole configuration file in attachement to be > sure that I didn't set something wrong that could replace or break > this confgiguration directive. > I could not say the symptons, sorry. > > Pavel > > > Andrew Hanushevsky wrote: > >> Hi Pavel, >> >> Could you give me the symptoms that made it appear to not work? The >> first >> one should have worked. If it dod not, this is a bug and will be fixed. >> >> In any case, a supervisor node needs *both* 'ofs.redirect remote' *and* >> 'ofs.redirect target' specified and it is subject to a redirection as >> well >> as redirecting itself. Specifying only "remote" will not work in most >> cases since the supervisor will not connect to the local olbd and off >> failures will occur (this pointed out that the protocol spec was a bit >> fuzzy on this, sigh). >> >> So, you should also have the line: >> >> ofs.redirect target if named supervisor >> >> Anyway, I will check on this. >> >> Andy >> >> P.S. I am in the process of simplifying the configuration which will >> make >> the "redirect" statement superfluous (though upwardly acceptable). >> >> On Tue, 28 Mar 2006, pjakl wrote: >> >> >> >>> Hi guys, >>> >>> I have deployed latest development version 20060307 and tried to >>> simplify my config file using INSTANCE_NAME and if statement. >>> I wrote this to the config file and for that particular node it >>> didn't work: >>> >>> if named dataserver >>> ofs.redirect target >>> else >>> ofs.redirect remote >>> fi >>> >>> If I did use like this, it did work: >>> >>> ofs.redirect target if named dataserver >>> ofs.redirect remote if named redirector >>> ofs.redirect remote if named supervisor >>> >>> (dataserver,redirector,supervisor) are INSTANCE_NAMES which I am >>> setting >>> through command line argument -n. >>> >>> Do am I missing something ? like badly comprehend the if statement >>> syntax ? >>> Thank for any advice >>> >>> Pavel >>> >>> >> > >------------------------------------------------------------------------ > ># !!!Removing this file is punished by death !!! # ># # ># This is a configuration file for XROOTD System # ># # > > # # > # XRD Daemon section # > # # > ># it specifies protocol and * means that library is statically linked with daemon >xrd.protocol xrootd * > ># it specifies xrd port (in future will be 1094 for manager and any for dataservers) >xrd.port 1095 > ># it specifies level of debuging ># --> normal mode ># xrd.trace all -debug -sched -poll -mem ># xrootd.trace all -debug -stall -mem ># --> debug mode, 2005/12 >xrd.trace conn net >xrootd.trace login > > # # > # XROOTD Section # > # # > ># here is loading the extended file system support for xrootd >xrootd.fslib /star/u/starlib/ROOT/xrootd/lib/libXrdOfs.so > >#here is loading support of authentication >xrootd.seclib /star/u/starlib/ROOT/xrootd/lib/libXrdSec.so > ># specify a valid path prefix to file requests >xrootd.export /data0 >xrootd.export /data1 >xrootd.export /data2 >xrootd.export /data3 >xrootd.export /home > > # # > # OLB Section # > # # > ># Restriction for hosts that subscribe to the olbd manager >olb.allow host rcas*.rcf.bnl.gov > ># Designate the port number to use for incoming requests from client (this is mandatory directive) >olb.port 3121 > ># Specify how load is computed and reported. >olb.perf int 5m pgm /star/u/starlib/ROOT/xrootd/utils/XrdOlbMonPerf.pl 300 > ># Specify the parameters for the load balancing scheduling algorithm >olb.sched cpu 20 io 40 mem 0 pag 10 runq 30 fuzz 10 refreset 43200 > ># Specify how server are selected for file creation >olb.space linger 5 500m 100m > > ># Designate paths handle by server (r - read, s will be staging) > >olb.path r /data0 >olb.path r /data1 >olb.path r /data2 >olb.path r /data3 >olb.path r /home ># with s stage all files from HPSS that aren't on distributed disk >#olb.path rs /home > >#Setup Manage Procesing latency >#@see documentation > >olb.delay discard 5 drop 10s full 5 lookup 1 overload * >olb.delay servers 90% service 60 startup 90 suspend 30 > > >#Set the time file existence information is to be cached in memory >#For testing .... >#olb.fxhold 60s > ># Tell server olbd's the location of the manager >olb.subscribe xrdstar.rcf.bnl.gov 3121 > ># Indecate the role this server will have based on host name (default is server) >olb.role manager if named redirector >olb.role server if named dataserver >olb.role supervisor if named supervisor > > >#Tell the odc component the location of the manager >odc.manager xrdstar.rcf.bnl.gov 3121 > ># Tell level of event loging >odc.trace all -debug >olb.trace all -debug > > # # > # OFS and OSS Section # > # # > ># ># Set redirect option using instance name (Supervisor must have both options) ># >ofs.redirect target if named dataserver >ofs.redirect remote if named redirector >ofs.redirect remote if named supervisor >ofs.redirect target if named supervisor > ># Restrict access to files only for operation read >oss.readonly > ># Tell level of loging >ofs.trace all -debug -aio -read -readdir -write >oss.trace all -debug > ># Control the way a disk cache partition is selected for file placement >oss.alloc * * 80 > ># Control file descriptor's allocating >oss.fdlimit * max > ># Specify whether or not to check that a file exists in HPSS >oss.check > ># Specify command to obtain data from HPSS >oss.mssgwcmd /star/u/starlib/ROOT/xrootd/utils/hpss_talk.pl >oss.stagecmd /star/u/starlib/ROOT/xrootd/utils/mps_Stage -F -c /star/u/starlib/ROOT/xrootd/config/configuration.cfg > ># Restrict access to files only for operation read >oss.readonly > ># Specify processing options for any entry matching the specified path prefix >oss.path /data0 >oss.path /data1 >oss.path /data2 >oss.path /data3 >oss.path /home >oss.localroot /home/starlib > ># Specify the location of a disk cache partition ># PJ: Only wildcard would work here. Individual drive would not and crash ># xrootd. >oss.cache rhstar /data* > ># number of threads, for beginning take defaults >oss.xfr 6 * 15 0 > > # # > # MPS section # > # # > ># Where error message go via mail, choose some user >mps.adminuser starlib > ># Specify path where will be placed symlinks >mps.scan /home > ># set level of debuging for all components (4 -additionally records internal debugging information) >mps.debug 4 > ># is the name of the file that is to hold error information >mps.errlogfn /star/data08/logerr > ># specify command for bringing the files from hpss >mps.stage.xfrcmd /star/u/starlib/ROOT/xrootd/utils/hpss_stage.pl %sfn %tfn > ># specify names of log files for every components of MPS separately >mps.stage.logfn /star/data08/stage_log > ># Set the number of times a failing data transfer operation may be retried. >mps.stage.max_retry 2 > >#is the name or number of the Unix group that is to be used for newly created files >mps.stage.cpgroup rhstar > >#is the command to be used to communicate with the Mass Storage System. >mps.stage.msscmd /star/u/starlib/ROOT/xrootd/utils/hpss_talk.pl > >#the command to issue in order to obtain free space information about a file system. >mps.stage._fs_stat /star/u/starlib/ROOT/xrootd/utils/fs_stat > ># enable security only one nodes that holds data >if named dataserver >sec.protocol /star/u/starlib/ROOT/xrootd/lib pwd -a:0 >fi >