[rdiff-backup-users] Re: Process hangs 100% with no strace activity

[Lorenzo Campanelli]

Hello,
  I just thought I'd let the community know that this had to do with common race condition bugs in libnss_ldap.so.2 regarding SSL and ldap.  Updating the system packages providing nss_ldap solved the problem.  It was only discovered after witnessing traditional tar processes having the same issue and looking at the coredump, of which LDAP lookups lock the process at the read() system call to the socket.  So no fault of rdiff-backup at all. Viva la rdiff-backup! :)

https://bugzilla.redhat.com/show_bug.cgi?id=222667
ps> This was on a standard CentOS 5 config that was badly in need of updating.
-Lorenzo

On Mon, Jun 1, 2009 at 9:55 AM, Lorenzo Campanelli <address@hidden> wrote:

Hello,
   I've been trying to troubleshoot an intermittent issue where a
rdiff-backup process gets stuck in an infinite loop with 100% CPU
cycles (shown via top).  When I check the processes using strace,
there is no activity.  I tried seeing if removing old slices would
help prevent this, but it comes back even when there are small amounts
of slice files.  Can this happen if the files are being
accessed/modified during backup?  Is this a sign of possible file
system corruption? My system log doesn't show anything regarding that,
but is this what could cause rdiff-backup to get stuck?  Have any tips
on troubleshooting this further?

-Lorenzo