Comment 1 for bug 281498

UPDATE: At 2 school sites, I have fixed the quoting issue to reflect the following:

---
if [ -n "$(ps ax|grep nbd$i)" ]; then
---

and went through the motions - rebuilt the image, rebooted thin-clients, etc. etc...

This seems to have helped the "Invalid operator" message upon TC bootup, but definitely has NOT fixed the issue of clients sitting at "Negotiation: " during nbd_swap file creation on the server. I see that 'dd' processes are running on the server, with little CPU time, yet I/O seems to be starved when multiple 'dd' processes are running simultaneously (such as when more than a few thin-clients are booted up at the same time, which is common practice when you have a lab full of thin clients, booting them all up at the beginning of the school day). They seem to sit for 2-4 minutes, depending on how many are being booted up at the same time. Because of the supposed I/O saturation, other server functions such as DHCP are unable to process, causing some thin-clients with flash memory to boot to that after failing to obtain an IP address via PXE/network boot.

I have tried what was suggested to me in #ltsp: a "elevator=deadline" I/O scheduler that is supposedly more suited for LTSP environments. This, after first test, reported "better" than the previous bootups, but still caused many machines to boot to flash because of not obtaining an IP address.

It would be greatly appreciated to have someone to help with this issue ASAP. It is critical to so many LTSP sites, that I'm wondering how anyone can reliably use nbd_swap functionality in Ubuntu. Thank you, and as always, I will help test any proposed solution/patch!

Cheers,
Jordan/Lns