I understood your servers run Lustre 2.7. Does robinhood run on a Lustre 2.7 client or an older version ? Did you build robinhood on a host where Lustre 2.7 is installed or an older version? Could it be possible that robinhood is currently processing changelog records that were generated while you was still running older Lustre (2.5?)
Also, it would be helpful to run "gdb" on a binary that include debug information, basically from the source tree: ./configure gdb --args ./src/robinhood/robinhood --readlog > run Thanks. On 06/15/15 19:50, Frederik Ferner wrote: > All, > > while upgrading our Lustre file system to Lustre 2.7, I also upgraded > robinhood to the newly released 2.5.5. I did download the tar file and > compiled it locally as the pre-built rpms on sourceforge have a > dependency on lustre-modules but on our site the rpm provides > lustre-client-modules. > > The RPM installed fine, the server is running Lustre 2.7 but with the > same configuration that previously was running fine (on 2.5.4) the new > version now segfaults on startup (called as robinhood --read-log). I'm > currently not sure how to debug this further. Any pointers welcome, > strace wasn't helpful in determining where it crashes, the log isn't > that clear either, with normal options the following are the only lines > in the logfile: > > <snip> > 2015/06/15 18:27:36 robinhood@cs04r-sc-serv-92[103117/1] CheckFS | > '/mnt/lustre03' matches mount point '/mnt/lustre03', type=lustre, > fs=cs04r-sc-mds03-01-10ge@tcp:cs04r-sc-mds03-02-10ge@tcp:/lustre03 > 2015/06/15 18:27:36 robinhood@cs04r-sc-serv-92[103117/2] SigHdlr | > Signals SIGTERM and SIGINT (daemon shutdown) are ready to be used > 2015/06/15 18:27:36 robinhood@cs04r-sc-serv-92[103117/2] SigHdlr | > Signal SIGHUP (config reloading) is ready to be used > 2015/06/15 18:27:36 robinhood@cs04r-sc-serv-92[103117/2] SigHdlr | > Signal SIGUSR1 (stats dump) is ready to be used > 2015/06/15 18:27:36 robinhood@cs04r-sc-serv-92[103117/1] EntryProc | No > class defined in policies, disabling file class matching. > 2015/06/15 18:27:36 robinhood@cs04r-sc-serv-92[103117/1] EntryProc | No > class defined in policies, disabling dir class matching. > 2015/06/15 18:27:36 robinhood@cs04r-sc-serv-92[103117/1] Main | Daemon > started (running modules: log_reader) > 2015/06/15 18:27:36 robinhood@cs04r-sc-serv-92[103117/3] ChangeLog | > LU-1331 is fixed in this version of Lustre. > </snip> > > With --log-level=DEBUG there are quite a few lines like this following > before it just stops: > > <snip> > 2015/06/15 18:46:19 robinhood@cs04r-sc-serv-92[110019/11] ChangeLog | > MDT0000: 3435143616 14SATTR 1434376871.895757017 0x14 > t=[0x20001026e:0x86f9:0x0] > 2015/06/15 18:46:19 robinhood@cs04r-sc-serv-92[110019/11] ChangeLog | > MDT0000: 3435143617 08RENME 1434376871.897757063 0x1 > t=[0x20000fa28:0xb3ef:0x0] p=[0xecedd2c:0x52385992:0x0] > LineScan$py.class s=[0x20001026e:0x86f9:0x0] > sp=[0xecedd2c:0x52385992:0x0] .LineScan$py.class.3FMhca > </snip> > > > Cheers, > Frederik ------------------------------------------------------------------------------ _______________________________________________ robinhood-support mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/robinhood-support
