Post - RC3 upgrade, clients don't stay logged in

Hi Shahid

I can't replicate this here, on either my UK timezone box or on my EST box.

You're sure there's not some underlying error - NIC, cable, switch - causing these dropouts. Note that it's normal for the clients to have their logged in light red in the management interface at times if their collection interval is greater than 300 seconds.

There's basically no change to the webservice side of things, other than the change of timezone (which is a one-line change), so I don't understand why the clients would suddenly disconnect all the time.

If you can't find an underlying problem, then you could try breifly reverting the server to 1.0.0-rc2. I think it should run OK with a 1.0.0-rc3 database (the changes made are all additions) but you would need to downgrade a client too.

Cheers

Alex

One other quick test, before doing any downgrade, would be to turn off the stats gathering (Management -> General -> Anonymous Statistics) if you have that turned on.

It hooks in to the requiredFiles call to the webservice, and might cause problems if your server doesn't have access to the internet (it shouldn't, I'm speculating). I don't see how that would cause problems registering clients though - it doesn't touch that routine.

Cheers

Alex

Server and client are on the same switch/vlan with the only changes being the upgrades on both. In trying the idea above about the stats gathering, I noticed that there are now 2 instances of the stats fields. My fresh install was a complete removal of the old RC2 xibo folder on the webserver and replacement with the new RC3 folder.

Oh, I have a screen cap of the above mentioned screen if you need it.

Sure. Attach it so I can see!

Sure. Attach it so I can see!

Could you just clarify what you mean by "I noticed that there are now 2 instances of the stats fields"?

Sure:

http://img.photobucket.com/albums/v223/omegagenesis/Xibo_Dual_Stats.jpg

(didn't see anywhere for attachments like in bug reports...)

Ta. That's very strange. And you say the upgrade ran OK?

It looks to me like the SQL got run twice.

Can you get me the result of the following SQL query please?

select * from `setting`;

What do you have in the client settings? Make sure you click on "Save" before attempting to register the client.

Bah - sorry. Wrong thread!

No prob. output:

http://img.photobucket.com/albums/v223/omegagenesis/SQL-result---Xibo.jpg

I did end up running the upgrade twice on the same database. But the problem existed after the very first upgrade install. Looks like there are some duplicate fields. I'll pour through it all and eliminate the duplicate entries. I might just have to nuke this DB and start over.

You should just be able to delete those last 4 rows from the setting table.

The only other major SQL change in the upgrade was adding a default layout. You might want to check the layout table to be sure there's no duplicates in there either.

Cheers

Alex

No joy. I'm sure I hocked something on my end. As I've only got the one client and 3 pieces of media to reupload, I'll just nuke the dbase and do another fresh install of rc3.

OK, another day... Nix'd the Xibo dbase and xibo web folders. For S&G's, put both the server and the client pc on a local network switch (little 5 port netgear) and verified communications between client/server. Re-ran install and got a fresh database. Initial registering of client went fine and it downloaded the new default layout. Set up a few test layouts and scheduled them for the display. Client did not check in for an update and the new layout never started. At that point, I tried to register a 2nd client and got the same error message below.

Tried to re-register the original client with same results, "unable to communicate with server".

http://img.photobucket.com/albums/v223/omegagenesis/Clienterror.jpg

I'm backrev'ing to RC2 for the time being until one of you brilliant minds can get back with me :)

Hey Shahid, sorry it took me this many days to join in the conversation I have been away on business.

If you have already done any of the below I apologise:

1. On RC3 navigate directly to http://yoursever/xibo/xmds.php - does the page render a description of the web service?
2. Enable Debug and Audit from Management -> Settings -> Error
3. Attempt to register a client
4. Navigate to Management -> Log. Filter for last 200 seconds (or since you clicked register). You can also filter on "xmds" from the page selection. Is there anything listed?

If there is nothing listed I suspect there is a fatal PHP error going on that we haven't checked. In this case go to /xibodir/lib/include.php and edit line 25

ini_set('display_errors', 0);

change to

ini_set('display_errors', 1);

-Re-navigate XMDS and see what occurs
-Re-register the display from client options

Sorry there is so much here to debug there is a blueprint that I really want to work on (https://blueprints.launchpad.net/xibo/+spec/server-mode) which would make this much easier I think!

Cheers,
Dan

1. It does, if it's supposed to display "view the WSDL for the service...etc.." and has a few links for register display/required files/etc...
2. check
3. check (same error)
4. zippo listed

Performed the next steps and still nada. The client, after registering once, pretty much refuses to connect to the server. My rc2 distro works just dandy with the exception of the prior scheduling issues.

Shahid,

This problem really has us stumped. Thanks for your patience and help! I'm going to try and add some more debugging information to the SOAP service for you - and either present that through the client results window if possible - or in the server log.

Thanks,
Dan

Still open pending extra debugging from Xibo SOAP service.

Thanks Dan, just let me know what I need to do to provide more information.

Shahid

Shahid

FYI, Dan and I met on Saturday and I setup my America/New_York server and client to try and reproduce this, but we couldn't.

It ran fine every time we tested.

Dan can get some more debugging from the framework that builds the SOAP service (NuSOAP) so once that's exposed to you, hopefully that will have some clue as to why this is happening.

We also went through the diffs from rc2 to rc3 and can't see anything that would cause the symptoms you describe - which is very strange.

Cheers

Alex

I have linked to a bug report to track the implementation of this extra logging from the SOAP service.

https://bugs.launchpad.net/xibo/+bug/348077

Once extra logging is available we can continue...

Extra info:

I just found out that if I delete the xlf files in the clients Xibo Library folder that it connects back to the server and loads whatever is next in the schedule.

Actually guys, I think we can go ahead and close this. I moved the server to a Win2k3 platform and segmented it into it's own vlan with static ip's and it has been running like a champ...

Great - thanks for letting us know.

There's a fix in 1.0.0 for the little "Logged in" traffic lights which means they should better reflect the state of the display (not that it makes any difference to the operation!).

Cheers

Alex