Re: [BackupPC-users] Memory leak prevents restore of large data sets.

[Craig B. <cba...@us...>]

John Pettitt writes:

> BackupBC running in a FreeBSD 5.4 box with windows clients.   Restoring
> a large (~200GB) data set to an smb share on the same FreeBSD box ...
> 
> Running: /usr/local/bin/smbclient \\\\jeeves\\Archive -U jpp -E -N -d 1 -c tarmode\ full -Tx -
> Running: /usr/local/backuppc/bin/BackupPC_tarCreate -h john-archiveone -n 40 -s hDrive -t /Exports /Stills
> Xfer PIDs are now 14226,14227
> [ skipped 203 lines ]
> tarCreate: Error: padding /data2/backuppc/pc/john-archiveone/36/fhDrive/fExports/fprofligate2/fB89S7423.TIF to 9789664 bytes from 9437184 bytes
> [ skipped 86 lines ]
> tarCreate: Out of memory during "large" request for 1052672 bytes, total sbrk() is 536137728 bytes at /usr/local/backuppc/lib/BackupPC/FileZIO.pm line 192.
> Empty file, short tar file, or read error: Unknown error: 0
> This backup will fail because: Abandoning restore
> Abandoning restore
> Restore failed: Abandoning restore
> 
> Watching the process with top it seems to grow by the second until the
> system eventually barfs the memory request and it dies.    This look
> like a memory leak to me.

Yes, it does look like a memory leak.  First, what version of
BackupPC are you using - I would like to know exactly where line
192 is.  Also, what version of perl are you using?

It appears to be something in the file uncompress code.  Perl
should handle memory allocation and freeing, so unless my code
has a detached circular hash reference there shouldn't be
leaks.  The other place to look is Compress::Zlib.  It's in
c code and might not clean up correctly (or BackupPC does something
that causes it to not clean up).  For that matter, what version
of Compress::Zlib do you have?

Craig