Re: [GNUnet-developers] Hang initializing fs

[Milan]

Christian Grothoff a écrit :
> The first time (this particular version of) gnunetd starts up (with sqlite 
> datastore) after an update, it needs to determine the actual size of the 
> datastore's entries (not just the filesize).  For that, gnunetd does a full 
> scan of the sqlite database.  If you abort gnunetd during that time, it will 
> restart the scan next time.  Until that scan is complete, the file-sharing 
> subsystem won't be up and running.  
>   
I finally got out of this mess by waiting. Maybe I would be useful to
add a debug level log saying that this check is starting and that it has
finished, because for now you can imagine that is a long-run behavior of
gnunetd.

Was that intended to be run only one time after an upgrade? Since I
removed everything in /var/lib/GNUnet (letting /etc/gnunetd.conf), it
looks like the check is performed from an empty database too. So maybe
it would definitely be good to tell it to the user.

> Depending on your database size and IO capabilities, that scan can take quite 
> a while.  The stacktrace that you are reporting below shows gnunetd doing the 
> scan.
>   
But even with a completely empty /var/lib/GNUnet/, gnunetd is seeking
for several minutes, even if I can't imagine what it can read/write
since the directory is only 4KB after some minutes of intense work.
That's really strange.

> I hope this helps...
>   
Sure, thanks. Hope this may help too.


PS: I'm currently working on our libgksu2 use in gnunet-gtk to make it
comply with new options (since now you can choose a custom config file
from the UI). I'll give you infos soon.