copy from ftp:// uses 100% CPU

Bug #29260 reported by Andrew Jorgensen
10
Affects Status Importance Assigned to Milestone
nautilus (Ubuntu)
Invalid
Medium
Ubuntu Desktop Bugs

Bug Description

I opened an anonymous ftp server with nautilus and copied a large (2GB) file to my local disk. During the entire transfer the CPU was at 100% with nautilus.

Revision history for this message
Lakin Wecker (lakin) wrote :

Hi Andrew,

Thanks for your bug report. I Just finished testing this on Ubuntu Dapper, and am unable to reproduce the bug.

Can you provide the version of Ubuntu and the version of nautilus that you are using.

If you are able, it would be great if you could test with dapper using the same FTP site/file. Also, can you provide the address for the public FTP site that you used and the file that was download. So that we can try to reproduce this bug.

I'm marking the bug as needs info, feel free to re-open it when you have provided the necessary information.

Changed in nautilus:
status: Unconfirmed → Needs Info
Revision history for this message
Andrew Jorgensen (ajorg) wrote :

Processor is sitting around 80% on an ftp copy using nautilus (X11 is around 6% according to top). This is a transfer of a 2GB file (DVD iso image) from an ftp site on the LAN (network settles at about 83% according to the system monitor applet).

80% is better than the 100% I was getting, but it's still not very good.

Then once it gets to 100% on the transfer (0:00 Remaining) the dialog sits there a very long time with nautilus taking about 27% and X11 taking about 11%.

By comparison curl and wget use around 40% CPU for the same operation (at around 100% network).

(all tests on dapper, updated before testing)

Changed in nautilus:
status: Needs Info → Unconfirmed
Revision history for this message
Andrew Jorgensen (ajorg) wrote :

The FTP server I'm connecting to is behind our firewall, but it's running NcFTPd. The fact that curl and wget use about half the CPU for the same operation leads me to believe that the server is not (entirely) to blame.

It may be that this bug only shows up if the transfer is very fast (local LAN).

Revision history for this message
Sebastien Bacher (seb128) wrote :

How fast is your network? Using a 100M local one nautilus uses around 4% of CPU to copy from a proftpd ftp server on my box using dapper

Changed in nautilus:
assignee: nobody → desktop-bugs
status: Unconfirmed → Needs Info
Revision history for this message
Andrew Jorgensen (ajorg) wrote :

The network is 100Mbps. My CPU is a 1.7GHz Pentium 4. Could the NcFTPd be causing the poor performance? I will try setting up a proftpd server at some point.

In the mean time is there some other way I can help figure out what's causing this? I'm not very handy with a profiler (never actually used one) but I can follow directions.

Revision history for this message
Sebastien Bacher (seb128) wrote :

I'm not sure of what could be useful, but wget using 40% of the CPU doesn't seem to be a normal CPU load. As said it uses around 4% to download on my box by example. If you could try on a another server just by comparaison it would be nice

Revision history for this message
Andrew Jorgensen (ajorg) wrote :

Installed proftpd on another box and tried pulling a file from it. Behavior there is basically the same. I also tried taking out user_xattr from the mount options and enabling the various performance tuning stuff on the HDD. No significant change.

Andrew Jorgensen (ajorg)
Changed in nautilus:
status: Needs Info → Unconfirmed
Revision history for this message
Sebastien Bacher (seb128) wrote :

If somebody who gets the issue want to track what happens maybe getting a gdb backtrace of what it is doing could be useful. Using sysprof would give the details of where it's spending cpu

Revision history for this message
Martin Bergner (martin-bergner) wrote :

Your bug lacks information we would need to investigate further. We
are now going to close the bug - please reopen if you have more
information at hand.

Changed in nautilus:
status: Unconfirmed → Rejected
Revision history for this message
Andrew Jorgensen (ajorg) wrote :

Closing this bug is fine by me at this point. The problem still exists but is nowhere near as bad as it was when the bug was filed.

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.