Error trying to save configuration

21 views
Skip to first unread message

tom...@gmail.com

unread,
Sep 5, 2008, 11:39:28 AM9/5/08
to Disco-development
Hi,

I'm trying to install and run disco locally. I can get homedisco to
work, I can get the server to run and the two lighttpd instances. I
can get to the config page. But when I try to save a new host row I
get "Request Failed".

Tailing the disco log reveals:

=INFO REPORT==== 5-Sep-2008::11:21:33 ===
Error{EXIT,{badarg,[{erlang,list_to_integer,[[]]},
{disco_config,parse_row,1},{disco_config,-update_config_table/1-lc
$^0/1-0-,1},{disco_config,update_config_table,1},
{disco_config,save_config_table,1},{handle_ctrl,handle,2},
{scgi_server,handle_request,2},{scgi_server,scgi_worker,1}]}}

I'm trying to add localhost, which is definitely in my /etc/hosts file
as 127.0.0.1. I also tried adding a "testhost" entry with my local ip
and adding that, which also failed with the same error.

If I try to add the host as "127.0.0.1" it fails with a different
error

=INFO REPORT==== 5-Sep-2008::11:32:26 ===
Error{EXIT,{function_clause,[{disco_config,add_nodes,[[],1]},
{disco_config,-update_config_table/1-lc$^0/1-0-,1},
{disco_config,update_config_table,1},{disco_config,save_config_table,
1},{handle_ctrl,handle,2},{scgi_server,handle_request,2},
{scgi_server,scgi_worker,1}]}}

Can anyone help?

Thanks!

Edmund

tom...@gmail.com

unread,
Sep 5, 2008, 12:14:30 PM9/5/08
to Disco-development
Update....if I modify the call to list_to_integer in disco_config.erl
as follows

27c27
< add_nodes(string:tokens(NodeSpec, ":"),
list_to_integer("1")).
---
> add_nodes(string:tokens(NodeSpec, ":"), list_to_integer(Instances)).

I don't get an error when saving, but returning to the page indicates
that the 1 instance setting has been lost. Looks like the instances
setting isn't getting translated correctly between the front and back
ends, like maybe the names in the js don't correspond to what's
expected in the erlang? Is anyone else having this problem?

tom...@gmail.com

unread,
Sep 5, 2008, 1:37:17 PM9/5/08
to Disco-development
Further update, it's a UI issue apparently.

The actual value for the max workers isn't sent back to the server
from my version of firefox at least.

The post body is [[" localhost",""]] whatever number I enter for max
workers.

Now that I know it's not a configuration error on my part I'll report
a bug.

tuulos

unread,
Sep 5, 2008, 3:51:03 PM9/5/08
to Disco-development

> Now that I know it's not a configuration error on my part I'll report
> a bug.

Thanks for the bug report.

Which version of Firefox you're using?

There's a known issue with the config table currently which is that
cells must be deactived (not gray) before clicking "save table".


Ville

Justin

unread,
Sep 9, 2008, 12:13:26 AM9/9/08
to Disco-development

I also get "Request Failed" when trying to add localhost. I've tried
safari 3.1.2 and firefox 3.0.1.

=INFO REPORT==== 8-Sep-2008::23:07:23 ===
Error{EXIT,{badarg,[{erlang,list_to_integer,[[]]},
{disco_config,parse_row,1},{disco_config,-update_config_table/1-lc
$^0/1-0-,1},{disco_config,update_config_table,1},
{disco_config,save_config_table,1},{handle_ctrl,handle,2},
{scgi_server,handle_request,2},{scgi_server,scgi_worker,1}]}}

Thank you for any insight,
Justin

matteo.acerbi

unread,
Sep 9, 2008, 6:50:54 AM9/9/08
to Disco-development
On 9 Set, 06:13, Justin <jta...@gmail.com> wrote:
> I also get "Request Failed" when trying to add localhost.  I've tried
> safari 3.1.2 and firefox 3.0.1.
>
> [..]

Getting similar responses with disco's 0.1 debian packages on a
mostly up-to-date debian sid with both firefox 3.0.1 and
konqueror 3.5.9:

Error{EXIT,{badarg,[{erlang,list_to_integer,[[194,160,50]]},
{disco_config,parse_row,1},{disco_config,-update_config_table/1-lc
$^0/1-0-,1},{disco_config,update_config_table,1},
{disco_config,save_config_table,1},{handle_ctrl,handle,2},
{scgi_server,handle_request,2},{scgi_server,scgi_worker,1}]}}

I'm just trying disco for the first time, if anybody could tell
me how to add a valid single node ({"localhost", 2}) bypassing
the web interface I'd be very grateful!

Of course, if I can help debugging why this UI doesn't work
here (but I'm not an expert in the field) just tell me what to
try, please.

Thank you.

Best regards,
Matteo

Matt Heitzenroder

unread,
Sep 9, 2008, 6:55:34 AM9/9/08
to disc...@googlegroups.com
Matteo-

I had a very similar issue with the web interface.

Once you you've added "localhost, 2"

Try clicking on localhost again to exit the number field. Then hit
save.

Let me know if that works.

Matt

Matteo Acerbi

unread,
Sep 9, 2008, 7:09:11 AM9/9/08
to disc...@googlegroups.com
On Tue, Sep 9, 2008 at 12:55 PM, Matt Heitzenroder
<mheitz...@gmail.com> wrote:
>
> Try clicking on localhost again to exit the number field. Then hit
> save.
>
> Let me know if that works.

I had already tried that and tried it again now, but unfortunately it
doesn't work here.

Apart from this problem, I can "load" count_words.py on the master,
but it won't run because of the lack of available nodes.

So I'd really like to know if there is a way to inform the master
process that there's an available dual-core node on localhost without
using this interface, it would be nice to see the example running
anyway. :-)

Thanks,
Matteo

Justin Lolofie

unread,
Sep 9, 2008, 8:48:07 AM9/9/08
to disc...@googlegroups.com

I'm no javascript expert, but the problem seems to be in config.js send_table(). When I try to add localhost with 2 workers, the jquery selector finds our tr. The tr has two td child nodes. The value of the first td is "&nbsp;localhost". The value of the second td is an html form. That doesnt seem right.

Anyway, if anyone is looking for a quick work around they could probably edit this line:
        jsonTable = ("[" + $.makeArray(table).join(",") + "]");
and change jsonTable to look the way they want it to: ["localhost","2"] maybe?

I'm interested in fixing the actual problem though- so if there are any javascript experts out there... :>

oh, also, if someone could fix this line in config.html:
        <tr><td> </td><td> </td></td>
The last tag should be </tr> ? (I dont think this is related to my problems.)

thank you,
justin

Justin Lolofie

unread,
Sep 9, 2008, 9:06:06 AM9/9/08
to disc...@googlegroups.com
Sorry, I worded this wrong. The problem isnt IN send_table(), its observable from there. :)

The real problem is somewhere else. I'm looking in jquery.uitableedit.js at the bit of code that is inserting a html form into a td.

-justin

mies

unread,
Sep 9, 2008, 10:10:08 AM9/9/08
to Disco-development
Hi,
I have a "disco_4444.config" in the "/srv" directory that contains the
information that is stored through the web interface. You can edit
this file to inform disco you have a dual core node available on the
localhost.

Hope this helps,

grtz
mies

Justin Lolofie

unread,
Sep 9, 2008, 10:54:33 AM9/9/08
to disc...@googlegroups.com

Ah, I love files. Thank you,
Justin

Justin

unread,
Sep 9, 2008, 11:47:10 PM9/9/08
to Disco-development

Is this what I want in disco_4444.config for one dual core node?
["localhost","2"]

When I try to run the count words example script, the httpd process
starts up but there is no mention of beam.smp. I've allowed STDOUT and
STDERR and this is what I get:

$ bin/start-master
Disco master started.
$ heart_beat_kill_pid = 13111
{"Kernel pid
terminated",application_controller,"{application_start_failure,disco,
{shutdown,{disco_main,start,[normal,[]]}}}"}
heart: Tue Sep 9 22:36:20 2008: Erlang has closed.
heart: Tue Sep 9 22:36:21 2008: Executed "". Terminating.

I'm not sure where to find out what actually went wrong. Thanks for
any help,
justin

Valentino Volonghi

unread,
Sep 10, 2008, 1:39:57 AM9/10/08
to disc...@googlegroups.com
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Sep 9, 2008, at 8:47 PM, Justin wrote:

>
>
> Is this what I want in disco_4444.config for one dual core node?
> ["localhost","2"]


That file format is JSON, what you want is this:

[["localhost","2"]]

- --
Valentino Volonghi aka Dialtone
Now running MacOS X 10.5
Home Page: http://www.twisted.it
http://www.adroll.com

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Darwin)

iEYEARECAAYFAkjHXa0ACgkQ9Llz28widGXTngCgw8k2NrO0JEfU24FKuIOmfpJ8
NHsAnjMo8t+J0mlXCRmvN+ywW1ir0N5l
=vb9H
-----END PGP SIGNATURE-----

Matteo Acerbi

unread,
Sep 10, 2008, 8:38:32 AM9/10/08
to disc...@googlegroups.com
On Wed, Sep 10, 2008 at 7:39 AM, Valentino Volonghi <dial...@gmail.com> wrote:
>
> That file format is JSON, what you want is this:
>
> [["localhost","2"]]

Thanks everybody for all of your suggestions, and to you in particular
for this one: I added that line to /srv/disco/disco_4444.config, fixed
the ssh authentication for user disco as explained at
http://discoproject.org/doc/start/install.html#configure-authentication
and this time everything worked fine with count_words.py.

When I have more time (unfortunately, not soon) I'll try and add some
other nodes and do some different experiments.

Best regards,
Matteo

PS: Will the debian repository be kept updated with the development
snapshots or will it deliver the stable releases over time?

tuulos

unread,
Sep 11, 2008, 5:42:49 AM9/11/08
to Disco-development

> Thanks everybody for all of your suggestions, and to you in particular
> for this one: I added that line to /srv/disco/disco_4444.config, fixed
> the ssh authentication for user disco as explained athttp://discoproject.org/doc/start/install.html#configure-authentication
> and this time everything worked fine with count_words.py.

Great! I'll check the Javascript issue with some recent Firefox.

> PS: Will the debian repository be kept updated with the development
> snapshots or will it deliver the stable releases over time?

I think we will setup a separate "unstable" repo for development
snapshots at some point.


Ville

Matteo Acerbi

unread,
Sep 11, 2008, 8:43:04 AM9/11/08
to disc...@googlegroups.com
On Thu, Sep 11, 2008 at 02:42:49AM -0700, tuulos wrote:
>
> Great! I'll check the Javascript issue with some recent Firefox.

Good, thanks. :-)

> > PS: Will the debian repository be kept updated with the development
> > snapshots or will it deliver the stable releases over time?
>
> I think we will setup a separate "unstable" repo for development
> snapshots at some point.

Oh, that would be wonderful in order to deploy the latest code easily
on the slave nodes.

On my development machine (where I'm running the master) I think I
will soon switch to the git version, though: the possibility of easily
modifying the code is of great help when trying to understand how it
works.

For example, when I said that the thing worked for me it wasn't true
at all :-) : the reduce phase can't really access the pairs produced
by the map. I guess the problem concerns the parsing of "dir://"
internal urls, but I'm going to try and debug this problem in depth
when I have more time, after I set up the git code (surely not earlier
than next week).

Best regards,
Matteo

Reply all
Reply to author
Forward
0 new messages