|
From: Martin J. <mar...@gm...> - 2012-05-16 10:11:27
|
Hi, few days ago I've upgraded my box from openssh-5.9 to openssh-6.0 and since then I've seen many hangs from backuppc (actually no full backup was completed, e.g. still running after 24hours, usually it finishes in ~ 3 hours - 60GB) rsync --server was running fine (I was able to rsync whole disk with it) perl process was stuck in select() call: pid 5553: /usr/bin/perl /usr/bin/BackupPC_dump -v -f 127.0.0.1 # strace -p 5553 Process 5553 attached select(16, [10], NULL, [10], NULL^CProcess 5553 detached # gdb --pid=5553 (gdb) bt #0 0x00007fb20ba3f6e3 in select () from /lib64/libc.so.6 #1 0x00007fb20cb69465 in Perl_pp_sselect () from /usr/lib64/libperl.so.5.14 #2 0x00007fb20cb1f306 in Perl_runops_standard () from /usr/lib64/libperl.so.5.14 #3 0x00007fb20cac0d6a in perl_run () from /usr/lib64/libperl.so.5.14 #4 0x0000000000400e89 in main () Usually it rsynced 24MB to new backup, sometimes more, but e.g. XferLogLevel and --verbose haven't changed that. Looking at openssh changelog I suspect this 2 changes: https://bugzilla.mindrot.org/show_bug.cgi?id=1859 https://bugzilla.mindrot.org/show_bug.cgi?id=1943 but it could be something completly different, but now after downgrade back to 5.9 I've rsynced 2,7G to new sofar and it's still running. I'll try to narrow this a bit more (even bisect openssh if needed), but wanted to report it here asap so people don't need to debug perl and other stuff like I did (not expecting openssh to be the cause). Or maybe someone already knows about better work around with -W or something which should be added to default BackupPC config.. Cheers, |