[Libguestfs] [PATCH] RHEV: Use dd and direct io to write to NFS

I've been experiencing severe stability issues writing large amounts of data to an NFS export (I have never successfully written 8GB of data without having to reboot my machine). This patch alleviates the problem. I have successfully exported an 8GB disk with this patch in place. --- lib/Sys/VirtV2V/Target/RHEV.pm | 29 ++++++----------------------- 1 files changed, 6 insertions(+), 23 deletions(-) diff --git a/lib/Sys/VirtV2V/Target/RHEV.pm b/lib/Sys/VirtV2V/Target/RHEV.pm index 4b663ef..9b4b73a 100644 --- a/lib/Sys/VirtV2V/Target/RHEV.pm +++ b/lib/Sys/VirtV2V/Target/RHEV.pm @@ -285,29 +285,12 @@ sub open path => "$path.meta", error => $!))); - # Open the data file for writing - my $data; - open($data, '>', $path) - or die(__x("Unable to open {path} for writing: {error}", - path => "$path", - error => $!)); - - # Write all data received to the data file - my $buffer; - - for(;;) { - my $ret = sysread(STDIN, $buffer, 64*1024); - die("Error in NFSHelper reading from stdin: $!") - unless (defined($ret)); - last if ($ret == 0); - - print $data $buffer; - } - - close($data) - or die(user_message(__x("Error closing {path}: {error}", - path => "$path", - error => $!))); + # Write the remainder of the data using dd in 2MB chunks + # XXX - mbooth(a)redhat.com 06/04/2010 (Fedora 12 writing to RHEL 5 NFS) + # Use direct IO as writing a large amount of data to NFS regularly + # crashes my machine. Using direct io crashes less. + exec('dd', 'obs='.1024*1024*2, 'oflag=direct', 'of='.$path) + or die("Unable to execute dd: $!"); }); } -- 1.6.6.1