Hello all,

 

I am testing curl in my network landscape and I'm hoping someone could give some tips for my situation.  We have a business need to bring nearly 100G or more of data across a WAN via FTP.  The existing solution is a script using WinNT's ftp and managing many ftp sessions in parallel (quad-processor server).  In the end we just hope nothing fails and all the files arrive safely.

 

We're finding this isn't good enough and after researching curl, this may be my alternative.  I have a script in Perl now that manages many curl processes (via Win32::Process) to download several files.  This works great, however, I'm finding that curl aborts several times at random with the error message "(23)  Failed writing body".  There is over 300G free on the drive I write to.  If it matters I am using a UNC path to a remote server for file storage.

 

Unfortunately curl is not able to resume broken downloads when connecting to an IIS 4.0 FTP server.  The script must then restart failed downloads.

 

As a test, I have run plain FTP and curl side-by-side storing files to the same remote directory via mapped drive (since FTP can't handle UNC).  I have seen curl abort a couple times where FTP just keeps on going.  NT Performance Monitor verifies this activity.

 

Any ideas?

 

---

Shawn Poulson

Basis Admin, SAP America