Re: Deprecating NBD in favor of squashfs-over-NFS

On 5/17/19 11:15 AM, Rolf-Werner Eilert wrote:
> Thank you Alkis for this explanation.
> 
> What would be the effect of instability or no reconnections? Maybe we 
> experienced some and just don't know that NBD was the reason... Anyone 
> here with an example?

Google for this exact phrase:
ltsp site:https://github.com/NetworkBlockDevice/nbd/issues

You'll see a few of the issues I reported to NBD;
there are others too, but they don't have the "ltsp" word in the report.

Hangs, kernel panics, things working one time but not another, *may* be 
due to NBD, but of course they may be due to a lot of other reasons.
For example if you reboot the server while a client is in the login 
screen, NFS will freeze and CONTINUE after the LTSP server comes back 
up; NBD will crash, exit, and cause kernel panics later on.


> Ok, best of both worlds means, speed AND stability?

Yes. Actually, NFS may be faster:
`tar -cf /dev/null /usr` needed 2.5 secs on NFS, 5 secs on NBD.
I couldn't see any speed difference while opening firefox/libreoffice 
though.


> That would make sense. But you should make a decent script which changes 
> this without destroying other (maybe manually made) edits in the config 
> files. Do you think that is possible?

If you read the "longer explanation" paragraph in my previous mail, I 
mention there that the pxelinux.cfg/default file is regenerated each 
time, so it doesn't contain any changes made by users.


Cheers,
Alkis


_____________________________________________________________________
Ltsp-discuss mailing list.   To un-subscribe, or change prefs, goto:
      https://lists.sourceforge.net/lists/listinfo/ltsp-discuss
For additional LTSP help,   try #ltsp channel on irc.freenode.net